Marketing, Content & Communications

How to get hired as an SEO manager in 2026-27

Updated 5 October 2026 · 22 min read · by SkillSculptor

The short answer

SEO Manager is still a hired job in 2026, but the version of it that got hired for publishing volume against a keyword list is mostly gone. What employers buy now is diagnosis and delivery: someone who can work out why a site is not being retrieved or cited, get the fix through an engineering backlog, and explain organic performance to an executive who has already watched the traffic chart fall. There is no licence, no required degree and no certification that functions as a credential in SEO, so the gate is a case file: three to five documented engagements where you state the diagnosis, the change you made, who had to ship it, how you measured it, and what you got wrong. Generative engine optimisation (GEO) is scope added to this role rather than a separate career, and the candidates who win interviews in 2026 and 2027 are the ones who can say which parts of it are measurable (crawler access verified in server logs, structured data, entity consistency, third-party mention accuracy, product feeds, citation share against a fixed prompt set) and which parts are still hypothesis, instead of presenting a confident framework for a system nobody can audit.

Credential gate: noneNo licence, no governing body, no accredited degree, no required exam. Nothing in SEO works like an electrician's licence or a nurse's NCLEX. Certificates exist (Google's analytics skill badges, Semrush and HubSpot academies, Ahrefs courses, CXL's paid programmes) and take hours to a few weeks. They read as keyword filler on a resume, not as a qualification. Hiring managers screen on evidence of work.
What replaces the credentialA case file. Three to five engagements, each written as property shape, symptom, diagnosis and how you reached it, the specific change, who shipped it, how you measured, result, and the part you could not attribute. One of them should be a decline you managed honestly. It takes a weekend to write and most applicants do not have one.
Typical hiring loopRecruiter or founder screen (20 to 30 minutes, tool names and site scale), hiring manager interview (diagnosis and judgement), a practical exercise (live teardown or take-home audit with a presentation), a cross-functional conversation with engineering or editorial about how you get work shipped, sometimes a 30-60-90 plan, then references. Mid-market in-house runs two to four weeks; an agency can decide in a week; enterprise takes six to eight.
Five employer types, five jobsAgency and consultancy, in-house brand and ecommerce, publisher and media, B2B SaaS, and marketplace or enterprise platform. The vocabulary overlaps; the work does not. Ecommerce SEO lives in faceted navigation and product feeds, publisher SEO in editorial workflow and news surfaces, SaaS SEO in a content programme and comparison pages. Pick the variant from the posting's nouns before you write anything.
What changed by 2026AI Overviews and AI Mode absorb clicks on informational queries, so impressions can hold or rise while clicks fall. Google launched Search Generative AI performance reports in Search Console on 3 June 2026: impressions and pages in generative AI features, split by country, device and date, with no query, click, CTR or position data and no backfill before launch. Crawler access became a business decision taken at the CDN. Postings now name GEO, AEO or AI visibility, and titles bundling search with AI visibility have appeared.
Where to get a real pay numberThe US BLS has no SEO occupation. Specialist and individual-contributor SEO work falls under OES 13-1161 (Market Research Analysts and Marketing Specialists); people-managing or budget-owning SEO leadership falls under 11-2021 (Marketing Managers). Both publish national and metro percentiles annually with roughly a year of lag, and neither isolates search work. The better source for your level is live postings: pay-transparency laws in states including California, Colorado, New York, Washington and Illinois mean a large share of ads carry a band. Collect twenty for your title, level and market.
The two proof problems you must solveFirst, your last two years of traffic charts may point down through no fault of yours, so you need a way to show skill without showing growth. Second, AI answers are not fully attributable, so you need a measurement story that separates what you can evidence (crawler access in logs, impressions, citation share against a fixed prompt panel, revenue by page cohort) from what you can only estimate. Candidates who pretend either problem away get rejected by anyone competent.
Fastest realistic routes inAgency junior or executive role (most entry openings, lowest pay, fastest learning), an internal move from content, paid search, analytics, web development or ecommerce merchandising, or freelance for two or three small businesses while you build a case file. An owned property with a public change log substitutes for professional experience better in SEO than in almost any other marketing role, because the work is verifiable from outside.

Yes, SEO is still a job. The posting tells you which of five jobs it is

Start with the question that brought you here. Companies are still hiring SEO managers in late 2026. Listing counts get published periodically by search recruiters and by the SEO tool vendors, and they agree on a shape rather than on a number: fewer generalist mid-level openings, a tilt toward senior, lead and head-of titles, and AI-search language in a growing share of ads. If you want a figure, open a job board and count it yourself for your title and your market this week. Any number printed on a page like this one is stale before you read it. What matters for your application is what the tilt means: the role that got cut produced content volume against a keyword list and reported sessions. The role that survived owns diagnosis, prioritisation and the relationship with engineering and editorial.

The second thing to accept is that "SEO Manager" names at least five different jobs that share a vocabulary of about forty words and almost nothing else. The tooling differs, the stakeholders differ, the metric that gets you promoted differs, and the interview differs. One resume sent to all five is the most common reason a competent SEO gets no replies.

Read the posting's nouns rather than its adjectives. "Shopify, Merchant Center, feed, faceted navigation, collection pages" is an ecommerce job. "Next.js, hydration, Core Web Vitals, edge, CDN, log files" is a technical job that will be tested technically. "Editorial calendar, briefs, freelancers, pillar pages" is a content job where your real skill is commissioning and internal linking. "Pipeline, MQL, demo requests, product-led" is B2B SaaS, where you will be asked to defend organic against paid on cost per opportunity. "Newsroom, Top Stories, Discover, syndication" is a publisher job with a different clock entirely.

Titles have also multiplied, and the title tells you the level of autonomy on offer. Treat the list below as a map: the same work gets four names depending on who wrote the job description.

What actually changed by 2026, and what did not

Be precise about this in interviews, because overstating the collapse and denying it both read as uninformed. The honest version: generative answers have taken a real and uneven bite out of clicks, concentrated on informational and definitional queries, while brand, navigational and high-intent transactional queries have held up far better. For most sites with a content footprint, organic search is still the largest single acquisition channel, and you should check that in your own analytics before repeating it about your own employer. The channel did not die. Its yield per impression fell, and the mix of what converts shifted.

The measurable consequence you will be asked about is the divergence between impressions and clicks. A page can be present in more results than ever, including inside AI surfaces, and earn fewer sessions. If your reporting treats clicks as the only proxy for visibility, you will describe a catastrophe where there is a mix shift, and you will have no way to show the work you did.

Google launched Search Generative AI performance reports in Search Console on 3 June 2026. Know exactly what they do and do not give you, because this question separates practitioners from people who read newsletters. They report impressions and the pages that appeared in generative AI features, broken down by country, device and date. They do not give queries, clicks, click-through rate or position for those surfaces, and the data starts at the launch window with no historical backfill. Google has said more metrics may follow without committing to a date, so check the current state of the report before an interview.

The second structural change is access. Whether AI crawlers may read your site stopped being a default and became a decision, taken at the CDN. Cloudflare's AI Crawl Control and its Content Signals syntax for robots.txt let a site allow, block, or charge for AI access, and the industry has moved from pay-per-crawl experiments toward usage-based arrangements. Publishers have signed licensing deals. This matters to your candidacy because somebody in the business has to hold an opinion on it, and increasingly that somebody is the SEO lead. If you cannot say what your current employer does about GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot, Google-Extended and Applebot-Extended, you are behind.

The third change is content economics. Drafting is effectively free, so volume is no longer a moat and in many categories it is a liability. What survives a quality review is first-hand evidence: original data, named and credentialed authors, product screenshots, tested claims, prices you actually checked. Both Google's systems and the answer engines are better at detecting consensus across independent sources than at rewarding any one site's assertions, which is why off-site mention work moved from a side activity to a core one.

The fourth change is where the answer gets assembled. For a growing share of commercial research questions, the sources an engine synthesises are third-party: review platforms, community threads, listicles and comparison sites rather than the brand's own pages. Part of the modern job is therefore earning accurate mentions on the pages answer engines retrieve, which looks more like digital PR and review-platform management than like on-page optimisation. Say this plainly in an interview. It is the part most candidates miss.

Where you can actually submit a site, and where there is no submit button

Job ads ask for AI visibility and candidates reach for theory. Knowing the real list of submission surfaces, and the places where no such endpoint exists, is a short piece of homework that makes you sound like someone who has done the job. It also lets you disqualify a vendor in one question.

Split the list into four groups: classic index submission, places, commerce feeds, and publisher programmes. Everything outside those groups is influence rather than submission, which means crawler access, content that is actually in the HTML, consistent entities, and accurate third-party sources.

For Google there is no submission API for ordinary pages. Your levers are an XML sitemap referenced in robots.txt and submitted in Search Console, internal links, and URL Inspection's Request Indexing for individual URLs, which is rate limited to a handful a day and is a diagnostic rather than a publishing workflow. Google retired the old sitemap ping endpoint in 2023, so any script still calling it is doing nothing. Google does not participate in IndexNow.

Bing is the exception worth real effort, and for a reason that is easy to explain in an interview. Microsoft Copilot's web results draw on Bing's index, so Bing Webmaster Tools and IndexNow carry more weight for assistant visibility than Bing's share of classic search would suggest. Bing Webmaster Tools gives sitemap submission, a per-site URL submission quota, a URL Submission API, and reporting on your IndexNow submissions.

For the assistants themselves there is nothing to submit. ChatGPT, Claude, Gemini and Perplexity have no URL submission endpoint for ordinary content pages, and no amount of schema markup creates one. Any service selling AI search submission or GEO directory listings for content pages is selling something that does not exist. It is the 2026 reissue of link-directory spam. The honest answer to "how do we get into AI answers" is the combination of the real feeds below, crawler access you can prove from logs, and getting cited on the third-party pages engines actually retrieve.

How SEO manager hiring actually works, and what each stage tests

This role has no licence, no accrediting body and no standard interview protocol, so the process varies more than in engineering or nursing. Four patterns cover most of it, and knowing which one you are in tells you what to prepare.

At an agency the process is fast and commercial. One or two conversations, often with the person who will bill you out, plus a short exercise. They are testing whether you can hold a client meeting, explain a drop without panicking, and carry several accounts. Sometimes they will offer a small paid trial project. Accept it if the scope is written down.

In-house at a mid-sized company the process is three to four stages over two to four weeks: recruiter screen, hiring manager, a practical exercise, and a cross-functional round with engineering, content or product. The practical exercise is where most candidates are lost, and it is almost always a variant of "here is a site, tell us what you would do".

At enterprise scale, add a panel, a take-home with a presentation, a skip-level conversation and a longer calendar, where six to eight weeks is normal. At a startup or a team-of-one hire, the founder may decide in two conversations, and the real test is whether you can describe the first ninety days in a way that sounds survivable without a content team.

One thing worth knowing about this field specifically: a meaningful share of good roles are filled through practitioner networks, newsletters, Slack and Discord communities and conference contacts before they are ever posted. The field is unusually public, people's work is visible, and hiring managers ask around. A quarterly habit of publishing a teardown, answering questions in a community, or speaking at a local meetup does more for your pipeline than another hundred applications.

Expect the screen itself to be keyword-literal. Recruiters without domain knowledge match on tool names, platform names, site scale and the title in the ad. If the posting says Screaming Frog, BigQuery and Shopify Plus, those words need to be on your resume in the context of real work, or a human may never see it.

The resume, and the case file that replaces a credential

An SEO manager resume is read twice: by a screener matching nouns, and by a practitioner looking for whether you made decisions. Serve both. The top third carries positioning plus the scale and type of site you have run; the bullets carry outcomes with the measurement named.

Scale and context are the fastest credibility signals in this field, and most candidates omit them. "Owned organic for a 400,000-URL marketplace across six locales on a headless stack, with two dedicated developers" tells a hiring manager more than any adjective. Always give the shape of the property: URL count, CMS or stack, locales, team size, and whether you had developer capacity.

Then the outcome problem. If your last two years of traffic went down with the market, do not hide it and do not present a chart that insults the reader. Write results that survive a declining channel: share of non-brand revenue, revenue or pipeline per page cohort, the gap between your result and a matched set of pages you did not touch, indexed-and-eligible page counts, time from recommendation to shipped fix, forecast accuracy, cost avoided. A candidate who can say "non-brand sessions fell against a category index that fell further, and non-brand revenue stayed flat because we re-weighted toward transactional templates" is more hireable than one claiming 300 percent growth with no baseline.

The bullets below are written as templates with brackets because the numbers have to be yours. Fill them from your own data. A bracket you cannot fill is a claim you do not have, and leaving it vague is worse than dropping the bullet.

Name the tools in context, once, in a line near the bottom, never as logos or as a five-star self-rating. The context is what matters: "log file analysis in BigQuery" beats "BigQuery" in a skills cloud.

Keep certifications to a single short line if you include them at all. Nobody is hired for a HubSpot badge, and a long certificate list on an SEO resume reads as a substitute for work. The exception is a genuinely technical credential that supports the variant you are applying for, such as a cloud or analytics engineering certification for an enterprise technical role. Two pages is fine past five years of experience, one page if you are early. The resume's job is to earn the exercise, not to contain your career.

Now the artefact that does the heavy lifting. Hiring managers know the market fell. What they cannot tell from a resume is whether you understood why, or whether you were simply present while it happened. The thing that answers that is a case file, and it is worth a weekend.

Write three to five cases, one page each, in a fixed structure: the property and its shape, the symptom as the business saw it, your diagnosis and how you reached it, the specific changes you made and who had to do them, how you measured, the result, and the part you could not attribute. That last field is what earns trust. Include one case that went badly or that you inherited mid-decline, and be concrete about the call you got wrong.

If your work is under NDA, anonymise rather than omit. "A mid-market B2B SaaS site, 40,000 URLs, three locales" is specific enough to be useful and vague enough to be safe. Never show a client revenue figure you were not given permission to share; percentages and indexed ratios are usually acceptable, and "redacted at client request" costs you nothing with a professional.

Add a forecast, because a written prediction compared against reality is one of the clearest separators in an interview. "I forecast the migration would cost fifteen percent of non-brand clicks for six weeks; the actual was twenty-two percent for nine, because I underestimated how long Google would take to re-evaluate the new category templates" is a sentence that gets people hired.

An owned property is the fastest substitute for professional experience in this field, because it is independently verifiable. Not a dead affiliate blog: a small site on a subject you actually know, with a public change log where each entry names a change, a date, a hypothesis and the measured outcome. Interviewers can look at it. Ten honest entries over six months beat a certificate. Add one artefact specific to 2026: a category prompt audit. Pick twenty to fifty questions a real buyer in your target employer's category would ask an assistant, run them, record which sources get cited, and write one page on the pattern. It takes an afternoon, demonstrates exactly the skill the job ad is gesturing at, and gives you something to open the interview with that is about their business rather than your history.

The interview: what it really tests, with the questions

SEO interviews reward a particular answer shape and punish another. The shape that works: hypothesis, evidence I would look at, decision I made, how I measured it, what I would do differently. The shape that fails is a list of best practices recited in order, and that shape is now actively suspicious, because any model can produce it on demand.

Expect to be given incomplete information on purpose. If an interviewer says "traffic dropped 40 percent, what happened", the correct first move is to ask what dropped: clicks or impressions, brand or non-brand, all templates or one, which countries, when exactly, and whether anything shipped that week. Candidates who name a cause in the first sentence ("sounds like a core update") lose.

Expect at least one question designed to see whether you will say "I do not know". Nobody can tell you how a given assistant ranked its sources. The honest answer, followed by what you would measure instead, scores higher than a confident framework.

Expect a question about engineering. In most companies the limiting factor on SEO is not knowledge but shipping capacity, and senior interviewers screen hard for whether you can get work into a sprint. Have a real story about a ticket that shipped and what made it shippable.

Finally, expect to be asked what you would not do. It tests seniority more efficiently than anything else, because it requires you to have been burned.

Measurement: how to defend your results while the chart falls

Read this section twice. It is both the hardest part of the job in 2026 and the thing interviews most reliably probe. The core move is to stop treating one number, organic sessions, as the proxy for the channel's health, and to replace it with a small set of segments that separate what the market did from what you did.

Build the brand versus non-brand split first, and get it agreed in writing before the quarter starts. Define the brand regex with the stakeholder who will read the report, because arguing about the definition after a bad month destroys credibility. Most of the apparent resilience or collapse in an organic chart is a shift in this mix.

Then segment by intent, and separately by whether generative features appear for the query. Pages whose queries trigger an AI answer behave differently from pages whose queries do not, and reporting them together makes both unreadable. Pull Search Console's bulk export into BigQuery, because the question "what happened to the subset of queries where an AI answer appears" cannot be answered in the Search Console interface.

Use the Search Generative AI reports for what they are: evidence of presence in AI surfaces, by page, country and date, without queries, clicks, position or CTR, and with no backfill before June 2026. They will tell you whether you are appearing. They will not tell you what the appearance was worth. Say both halves out loud in an interview.

Verify access in server logs. This is the one GEO input you fully control and can prove: whether GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot, Google-Extended, Bingbot and Applebot are reaching your important templates, at what rate, with what status codes, and whether your CDN or WAF is quietly blocking some of them. Plenty of sites discover a 403 pattern they never intended. Finding one is a legitimate, dated, defensible win you can put on a resume.

Stand up a prompt panel for citation share. Fix a set of fifty to two hundred prompts real buyers in your category would ask, run them on a schedule across the engines your customers actually use, and track how often your brand and pages are mentioned or cited versus named competitors. Tools exist for this (Profound, Semrush's AI visibility toolkit, Ahrefs' Brand Radar, Peec AI, Otterly and Scrunch among them). They are priced as monthly SaaS and the pricing moves, so check it when you build the business case. Know their limits before you present their output: answers are sampled and non-deterministic, so the same prompt yields different sources on different days; the prompt set is your guess at demand unless the vendor supplies real prompt data; and none of them can give you clicks.

Segment assistant referrals in analytics. Traffic from assistant surfaces is usually a small share of sessions and frequently converts above site average, because the visitor arrives already informed. Report it as its own channel with its own conversion rate from the first month, so that when it grows you have a baseline, and so that nobody can claim you missed it.

Then build the counterfactual, because that is what makes a falling chart defensible. Take the pages you changed and a matched set you did not, in the same templates with the same AI-feature exposure, over the same window. The difference between those two lines is the closest thing to your contribution this channel permits. Present it with the caveats attached. A hiring manager who has lived through 2025 and 2026 recognises it immediately as the work of someone who has had to defend a budget.

Finally, say what cannot be measured. There is no click data inside most AI answers, no reliable query data for AI surfaces, and no deterministic ranking to reverse-engineer. The professional position is to name the uncertainty, state what you are doing anyway because it is robust under either outcome (access, entity clarity, genuinely useful original content, accurate third-party mentions, clean feeds), and refuse to invent precision.

Pay, levels, and the contract versus permanent decision

There is no authoritative SEO salary table, and anyone quoting one precise national figure is guessing. Here is how to get a number you can defend in a negotiation.

Start with the BLS Occupational Employment and Wage Statistics. SEO work has no code of its own: individual-contributor and specialist roles map to 13-1161, Market Research Analysts and Marketing Specialists; people-managing or budget-owning roles map to 11-2021, Marketing Managers. Both publish national and metropolitan percentiles annually with about a year of lag. Use them for the shape of the distribution and the geographic spread, not as your target, because neither isolates search work and the manager code covers far broader jobs.

Then do the work that actually pays. Collect twenty live postings for your exact title, level and market, and record the band each one lists. Pay-transparency laws in states including California, Colorado, New York, Washington and Illinois mean a large share of ads carry a real band the employer has to honour. That is the best salary data available in this field, it is current, and it is free. Sort it by employer type, because the spread between an agency manager and an enterprise technical lead in the same city is large. While you are in the spreadsheet, tag each ad for whether it asks for AI visibility or GEO scope and compare the two groups yourself. That is a ten minute analysis on data you gathered, and it beats any premium figure quoted secondhand.

Structural patterns, stated as patterns rather than numbers. Agencies pay below in-house at the same level and compensate with breadth and speed of learning. Enterprise technical and platform SEO pays at the top of the range because the work borders engineering. Publisher SEO sits lower and is the most exposed to the current market. B2B SaaS pays well and attaches equity that may or may not be worth anything. Fully remote roles are increasingly banded by the employer's location policy rather than by yours.

Levels, in the order most organisations use them: specialist or executive, senior specialist, manager, lead or head of SEO, director of organic or search, VP of growth or marketing. Clarify two things in the interview, because titles lie: whether the manager title comes with direct reports or is a senior individual contributor role, and who owns the content budget. A manager with no budget and no reports is a specialist with a nicer title, which is fine if the pay reflects it.

Contract and freelance work is more viable in SEO than in most marketing roles, because audits, migrations and interim leadership are discrete products a buyer understands. The trade is predictability. Price monthly retainers rather than hours where you can, insist on a written scope for the first project, keep a minimum of three clients so losing one is survivable, and budget for self-employment tax, insurance and the unpaid business development that permanent employment hides. A common sequence: agency for two or three years to see many sites, in-house to learn depth and stakeholder work, then freelance or fractional with the case file already built.

One negotiation note specific to 2026. If a posting asks for AI visibility or GEO scope, that is new work being added to an existing role, and it is reasonable to negotiate on it: budget for a visibility tool, engineering time, and a clear statement of what success means in the first two quarters. A hiring manager who cannot answer "what would good look like at six months" is telling you they bought a buzzword and will judge you against a feeling.

Working with AI in this role

What an SEO manager has to know about AI in 2026-27

Of all the marketing roles, this is the one AI changed most, and it is still worth being precise, because the hype and the reality diverge. What changed is the surface where answers are delivered, the measurement that follows from it, and the economics of content production. What did not change is the craft underneath: whether a machine can reach your pages, parse them, resolve what they are about, and find your claims corroborated elsewhere. Crawlability, information architecture, internal linking, canonicalisation, rendering, page speed, entity clarity and earned authority still determine whether you are retrievable. GEO is not a replacement discipline. In practice it is classic technical and editorial SEO plus three genuinely new workstreams: access policy, mention coverage, and measurement of citation rather than clicks.

Say that out loud in interviews, because the opposite claim is now a red flag. There is no published ranking system for ChatGPT, Claude, Gemini, Perplexity or Copilot answers, no keyword tool with real volumes for prompts, and no click data inside most answers. Anyone selling a GEO methodology with the confidence of a 2015 ranking-factor study is guessing. The candidate who separates the evidenced from the hypothesised is the one a competent hiring manager trusts with a budget.

On the production side, treat drafting capacity as free and therefore worthless as a differentiator. Your competitors can publish as much as you can, instantly. What cannot be generated is first-hand evidence: a test you ran, a dataset you collected, prices you verified, a named expert's judgement, photographs of the actual thing. The content that both search and answer engines reuse disproportionately carries at least one of those. A hiring manager who has watched a competitor's generated library get pruned out of the index wants to hear that you know the difference between volume and substance, and that you have deleted pages as well as published them.

On the retrieval side, understand how an answer gets assembled at the level that is actually observable: a question is expanded into several related queries, passages are retrieved from a search index and from the model's own sources, and the answer is synthesised from what agrees. Two consequences follow. Passage-level clarity matters, because a self-contained paragraph that answers one question plainly is more reusable than the same information spread across a 3,000-word narrative. And consensus matters: if six independent sources describe your product one way and your site says something different, the engine follows the six. That is why a large part of modern GEO work is off-site, getting your category's review platforms, comparison pages, community threads and listicles accurate about you.

Knowing which submission surfaces genuinely exist separates practitioners from theorists, and the list is short. For classic indexation: XML sitemaps and URL Inspection in Google Search Console (Google retired the old sitemap ping endpoint in 2023, so scripts that call it do nothing), Bing Webmaster Tools, and IndexNow, which pushes changed URLs to Bing, Yandex, Seznam and Naver. Google does not participate in IndexNow, so for Google the levers remain sitemaps, internal links and crawl health. Because Copilot's web results draw on Bing's index, Bing Webmaster Tools and IndexNow matter more to assistant visibility than Bing's search share suggests, and Bing Webmaster Tools reports on your IndexNow submissions. For commerce: Google Merchant Center, Microsoft Merchant Center, OpenAI's merchant and product feed programme for shopping in ChatGPT, and Perplexity's merchant programme, which accepts a Google Shopping specification feed. For places: Google Business Profile, Apple Business Connect and Bing Places. For publishers: Google Publisher Center for News and Discover eligibility, plus any direct licensing arrangement your company has signed. Note what is not on that list: Preferred Sources is chosen by readers, not submitted to, and the assistants themselves have no URL submission endpoint for ordinary content. No amount of schema markup creates one.

The engineering-adjacent work is where the measurable GEO wins live. Decide and document which AI user agents may access what, at the CDN as well as in robots.txt, and know the difference between a training crawler and a retrieval crawler, because blocking the retrieval one removes you from answers while blocking the training one does not. Read your own logs for those agents. Make sure content is in the HTML rather than assembled client-side, because retrieval fetchers are less patient than Googlebot. Keep structured data accurate and keep organisation, person, product and article entities consistent across your site and your off-site profiles. None of this is speculative, all of it is verifiable, and it is the part of GEO you can put a date and a number against.

Finally, be honest about where AI did not change this job. The hiring process is still a human one. The metric that keeps your budget is still revenue. The reason most SEO recommendations fail is still that nobody shipped them. The daily skill that most separates a good SEO manager from a mediocre one is still the ability to persuade an engineer and a writer to do something they had not planned to do. A candidate who presents AI fluency as a substitute for that will lose to one who presents it as an addition.

Running a real prompt audit of your target employer's category, and knowing who gets cited and why

Every job ad now gestures at AI visibility, and almost every candidate answers with theory. Arriving with the actual answer landscape for the employer's own buying questions proves the skill and shows you did the homework. It also surfaces the uncomfortable truth most employers have not faced, which is usually that third-party sites and communities, not their own pages, own the answer.

Show it: Pick twenty to fifty questions a real buyer in their category would ask. Run them across the engines their customers use. Record which domains are cited, how the brand is described and whether the description is accurate, and write one page: who owns the answer today, which of those sources you could realistically influence, and the first three moves. Bring it to the first interview. If you have done this at work, state the scale and cadence: prompt count, engines, how often it ran, and one change it caused.

Owning AI crawler access policy end to end, including the CDN, and proving it from logs

This is the one GEO input you fully control, it is binary, and it is frequently broken by accident. Sites block retrieval crawlers while intending to block training crawlers, or a WAF rule silently returns 403 to the agents that would have cited them. Somebody has to hold the allow, block or licence decision, and increasingly that person is the SEO lead. Employers can verify this work, which is rare in GEO.

Show it: Name the agents you have made decisions about (GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot, Google-Extended, Applebot-Extended, Bingbot) and the difference between the training and retrieval ones. Say where you implemented it: robots.txt, Cloudflare AI Crawl Control or an equivalent CDN control, Content Signals, WAF rules. Then give the evidence: "found [measured] of retrieval crawler requests returning 403 at the edge, fixed in [timeframe], AI-surface impressions for the affected templates moved [measured]."

Measurement under click dilution: brand versus non-brand, intent cohorts, AI-feature cohorts, and an honest counterfactual

Traffic charts are falling across whole categories, so separating market movement from your contribution is now the core professional skill of the role, not a reporting chore. It is also what a CFO needs in order to keep funding you. Interviews probe it directly, usually with some version of "how would you prove this was worth it".

Show it: Describe the reporting you built, the tooling underneath it (Search Console bulk export to BigQuery, SQL, Looker Studio or a warehouse-native equivalent) and the decision it changed. Then describe one counterfactual you actually ran: the pages you changed, the matched set you did not, the window, the difference, and the caveats you attached when you presented it. Name what you could not measure. That sentence is the differentiator.

Knowing the real submission surfaces, and being able to say which ones do not exist

Employers are being sold AI visibility services, and the SEO lead is the person expected to tell them which are real. Knowing that sitemaps, URL Inspection, Bing Webmaster Tools, IndexNow, Merchant Center feeds, the ChatGPT and Perplexity merchant programmes, Google Business Profile, Apple Business Connect and Google Publisher Center are the actual endpoints, and that the assistants have none for ordinary content, turns a vague conversation into a budget decision. It also protects you from being handed an impossible mandate.

Show it: In the interview, answer "how would we get into AI answers" by naming the surfaces you can push to, the ones you can only influence, and the ones that do not exist. If you have run them, give mechanics: which feeds you own, submission or refresh cadence, whether IndexNow fires automatically from your CMS or CDN, and one thing you stopped paying for once you checked whether it did anything.

Entity and structured data work, done for machine resolvability rather than for rich results

Retrieval systems need to resolve what a page is about and who is making a claim. Inconsistent organisation naming, missing author identity, absent product and offer data and contradictory off-site profiles all make a site harder to reuse in an answer. The work is cheap, durable and robust whichever way the AI surfaces evolve, which makes it the safest investment you can argue for.

Show it: Give specifics: which schema types you implemented and validated, how you made organisation, person and product entities consistent across the site and off-site profiles, what you fixed in authorship and credentials, and what changed in eligibility or AI-surface impressions. "Rebuilt Product and Offer markup across [number] templates, cut Merchant Center disapprovals from [measured] to [measured], and added author entities with verifiable credentials across [number] pages" is the register.

Off-site mention and consensus work: review platforms, comparison pages, communities and digital PR

For a growing share of commercial questions the sources being synthesised are third-party. If the comparison pages, review sites and community threads in your category describe you inaccurately or omit you, no on-page change fixes that. This is the workstream most candidates have no story for, so having one is disproportionately valuable.

Show it: Describe a campaign with a measurable footprint: an original-data study and the publications that covered it, a programme to get accurate listings on the review platforms your buyers read, a corrections push on comparison pages that misstated your pricing, or a legitimate community presence with its rules respected. Report referring domains, the named placements, and movement in brand search demand or in citation share for the prompts you track.

Using language models inside the SEO workflow, with a stated quality line

Employers want to know you will not be a drag on tools they already pay for, and equally that you will not flood their site with generated pages. The distinguishing candidate uses models to delete work (clustering, log and query analysis, internal link candidates, brief construction, redirect mapping, QA at scale) and can say where they deliberately refuse to use them.

Show it: Give one concrete before and after with a measured figure: "clustered [number] queries into [number] intent groups and built the internal linking map in [timeframe] instead of [timeframe]", or "wrote a script that reconciles [number] log lines against the sitemap weekly, replacing a manual audit". Then name the refusal and the reason, for example no published page without a named human owner and a verified claim, and say what happened when someone asked you to do otherwise.

Product feeds and agentic commerce, for any ecommerce or marketplace role

Where buying happens inside an assistant, the feed is the surface. Merchandising data quality, price and availability accuracy, GTIN coverage and variant structure determine whether products are eligible at all, and the work is unglamorous enough that many teams have nobody owning it. For ecommerce employers it is the clearest line from your work to revenue.

Show it: Name the programmes you have fed (Google Merchant Center, Microsoft Merchant Center, OpenAI's ChatGPT merchant feed, Perplexity's merchant programme, which takes a Google Shopping specification feed) and the mechanics you own: feed generation and delivery, refresh frequency, attribute coverage, disapproval triage. Quantify: "raised GTIN coverage from [measured] to [measured] across [number] SKUs and cut disapprovals [measured], with [measured] incremental revenue attributed to the affected surfaces." Know the agentic checkout standards by name.

SQL and log analysis, so your answers do not depend on a vendor dashboard

Most of the questions that matter in 2026 cannot be answered in the Search Console or Google Analytics interface, because they require joining query data, crawl data, log data and revenue data. Candidates who can do that join are trusted with the strategy; candidates who can only read a tool's dashboard are managed. It is also the skill that makes AI-visibility claims checkable rather than anecdotal.

Show it: Say what you query and where: Search Console bulk export in BigQuery, server or CDN logs, the warehouse table that holds orders. Give one analysis that changed a decision, with the shape of the data: "joined [number] months of query data to order revenue by template and found [measured] of non-brand revenue concentrated in [number] page types, which moved the content budget." Mention Python or notebooks only if you really use them.

What a screen is looking for

These are the terms that a resume screen, human or automated, is matching against for this role. Use the ones that are true of you, in the words the posting uses.

Mistakes that cost people this job

Sending the same resume to agency, ecommerce, publisher, SaaS and enterprise platform SEO roles because they share a title.

Rewrite the top third for the variant. Ecommerce leads with revenue, templates, faceted navigation and feeds. Publisher leads with volume, speed, editorial workflow and news surfaces. SaaS leads with pipeline and comparison pages. Enterprise leads with URL counts, locales, crawl budget and the engineering relationship. Agency leads with accounts held, retention and pitching. Same career, five documents.

Opening with "AI killed SEO" fatalism, or with the opposite claim that nothing has changed.

State the specific, defensible version: clicks have been absorbed mostly on informational queries, impressions and brand and transactional journeys held up better, organic search is still the largest single channel for most content-bearing sites, and the measurement had to change. Then say what you did about it. Both extremes read as someone who has not looked at their own data.

Leading with "increased organic traffic 300 percent" with no baseline, no timeframe and no mechanism.

Give the base, the window, the cause and the counterfactual, in a line of this shape: "non-brand clicks from 42k to 96k over nine months, driven by consolidating 2,100 thin pages into 180 templates, while a matched set of untouched templates fell in the same window." Unanchored percentages are read as noise by anyone senior.

Presenting a GEO methodology as though AI answer ranking were a known system.

Separate the evidenced from the hypothesised. Evidenced: crawler access verified in logs, structured data validity, passage-level clarity, third-party mention accuracy, feed eligibility, citation share against a fixed prompt panel. Hypothesised: everything about internal weighting. Saying which is which is the senior signal.

Paying for an "AI search submission" or "GEO directory" service, or listing one as a win on your resume.

Know the real list and you can disqualify the vendor in one question. ChatGPT, Claude, Gemini and Perplexity have no URL submission endpoint for ordinary content pages, so nothing is being submitted. The surfaces that exist are sitemaps and URL Inspection in Google Search Console, Bing Webmaster Tools and IndexNow, Google and Microsoft Merchant Center plus the ChatGPT and Perplexity merchant feed programmes for products, Google Business Profile, Apple Business Connect and Bing Places for locations, and Google Publisher Center for news. Everything else is influence: access, clarity, corroboration.

Publishing an llms.txt file and listing it as AI visibility work.

Spend the time on things with evidence behind them. Google representatives have said publicly that Google does not use llms.txt, no major answer engine has committed to reading it, and you can check it yourself in five minutes by grepping your access logs for requests to the file. On most sites the count is zero or close to it. The file has a legitimate but different use: giving coding agents and documentation tools a clean map of a site. Mention it only with that caveat.

Bringing a 60-item audit output from a crawler to the exercise stage.

Bring three prioritised problems with estimated impact, estimated effort, how you would validate each, and one ticket written the way a developer would receive it. Then name what you could not assess without access to analytics and logs. Breadth is what the tool did; prioritisation is what they are hiring.

Not looking at the employer's site before the interview.

Spend ninety minutes: crawl a sample, check indexation of the money templates, run ten of their category's buying questions through two assistants, note who gets cited, and find one thing that is clearly broken. Lead with it. This single habit converts interviews in this role more reliably than any answer you can rehearse.

Listing every certificate, badge and course as though they were credentials.

Cut them to one line or omit them. There is no licence in SEO and no certificate functions as one. The space is better spent on the shape of the properties you have run: URL count, stack, locales, team and developer capacity.

Blaming the previous employer's developers, agency or leadership for the lack of results.

Own the delivery problem, because getting work shipped is the job. Describe what you changed in how you asked: writing requirements as tickets with acceptance criteria, attaching a revenue estimate, bundling SEO fixes into releases already planned, or building one relationship on the platform team. Interviewers are listening for whether you will be blocked here too.

Reporting rankings as the headline metric.

Report the chain: impressions and position to show presence, clicks to show capture, sessions and conversion to show value, and revenue or pipeline by page cohort to show the business outcome. Keep rank tracking as a diagnostic. An executive shown average position as a result has learned nothing they can act on.

Hiding or apologising for a decline you managed.

Present it as the case study it is: the decline, what caused it, what you tried, what worked, what you stopped doing, and the benchmark you measured against. Everyone senior in this field has lived through 2025 and 2026. The candidate who can narrate a decline honestly is more credible than the one with only up-and-to-the-right charts.

Treating off-site mention work as somebody else's job.

Take it on explicitly, because for many commercial questions the sources being synthesised are third-party. Audit how your category's review platforms, comparison pages, listicles and community threads describe you, get the inaccuracies fixed, and earn new coverage with something genuinely new. Report it in referring domains, named placements and citation share.

Accepting an AI visibility mandate with no definition of success, no budget and no baseline.

Negotiate it at offer stage. Ask what good looks like at six months, who else is accountable for brand mentions and PR, whether there is budget for a visibility tool and for engineering time, and what the current baseline is. If nobody can answer, you are being hired against a feeling, and that is worth knowing before you start rather than at your first review.

Questions people ask

Is SEO still a career worth entering in 2026?

Yes, with a changed shape. Organic search remains the largest single acquisition channel for most sites with a content footprint, and companies are still hiring SEO managers, but the market has tilted toward senior, lead and head-of roles and away from mid-level generalists who produced content volume against a keyword list. The work that is clearly still funded is diagnosis (why is this site not retrieved or cited), delivery (getting fixes shipped through an engineering backlog), measurement (separating market decline from your contribution) and AI visibility. The work that is genuinely shrinking is manual content production, rank reporting and audits nobody acts on. If you are entering now, go in through an agency, a team-of-one in-house role, or freelance work for small businesses, and build a case file from the first month.

What does an SEO manager actually do day to day?

Four things, in roughly this proportion. Diagnosing: working out from Search Console, analytics, crawls and server logs why performance moved, for which templates and which query segments. Persuading and shipping: writing requirements engineers will accept, getting them into a sprint, briefing writers, and chasing fixes to production. Measuring and reporting: maintaining the brand and non-brand split, the intent cohorts, the AI-surface view, and the revenue-per-page-cohort number that keeps the budget. Deciding what not to do: most SEO backlogs contain more work than any team can ship, so prioritisation is the actual skill. In 2026 add a fifth: owning the questions about AI crawler access, citation share and feed eligibility that nobody else in the business is positioned to answer.

Do I need a degree, certification or licence to be an SEO manager?

No. SEO has no licence, no governing body, no accredited qualification and no required exam. Most postings ask for a degree in marketing or a related field, and most hiring managers will ignore that line for a candidate with demonstrable results. Certificates from Google, Semrush, HubSpot, Ahrefs or CXL take hours to a few weeks and function as resume keywords at best; none is a credential in the way a nursing licence or an electrician's licence is. What substitutes for a credential is a case file of three to five documented engagements and, if you lack professional experience, an owned site with a public change log an interviewer can verify from outside.

Is GEO a different job from SEO, and should I apply for GEO roles?

GEO (generative engine optimisation, sometimes called AEO or AI visibility) is scope added to the SEO role, not a separate profession, and at most companies it is being handed to the existing search lead. It appears routinely in job descriptions now, and titles bundling search with AI visibility have started showing up, so apply for both with the same core evidence. The honest content of GEO work is this: make sure retrieval crawlers can access your pages and prove it from logs, make pages passage-clear and entity-unambiguous, get the third-party sources that answer engines synthesise to describe you accurately, run product feeds where commerce applies, and measure citation share against a fixed prompt panel. Anyone claiming to know how a given assistant ranks its sources is guessing.

Where can you submit a site to AI search engines?

For the assistants themselves, nowhere. ChatGPT, Claude, Gemini, Perplexity and Copilot's answer layer have no URL submission endpoint for ordinary content pages, and schema markup does not create one, so any service selling AI search submission or GEO directory listings for content pages is selling something that does not exist. The submission surfaces that are real: XML sitemaps and URL Inspection in Google Search Console (Google retired the old sitemap ping endpoint in 2023 and does not participate in IndexNow); Bing Webmaster Tools, which offers sitemaps, a URL submission quota and a submission API and matters for Copilot because Copilot's web results draw on Bing's index; IndexNow, which pushes changed URLs to Bing, Yandex, Seznam and Naver in one request; Google Merchant Center, Microsoft Merchant Center, OpenAI's merchant feed programme for shopping in ChatGPT and Perplexity's merchant programme for products; Google Business Profile, Apple Business Connect and Bing Places for locations; and Google Publisher Center for News and Discover. Preferred Sources is chosen by readers, not submitted to. Everything beyond that list is influence rather than submission: crawler access you can verify in logs, content that is in the HTML, consistent entities, and accurate descriptions of you on the third-party pages engines retrieve. Publishing an llms.txt file is not a lever: Google representatives have said Google does not use it, no major engine has committed to reading it, and your own access logs will show almost no requests for it.

How do I prove SEO results in an interview when organic traffic has been falling?

Replace the single traffic number with four things. First, segmentation: brand versus non-brand, by intent, and separately for queries where AI answers appear, so a decline can be located rather than mourned. Second, a benchmark: your result against a category index or a matched set of pages you did not touch, in the same window. Third, business outcome by page cohort: revenue, pipeline or ad yield, which is the number that keeps a budget. Fourth, delivery evidence: recommendations shipped and time from recommendation to production, which is entirely within your control. The sentence that lands has this form: non-brand sessions fell against a category that fell further, and non-brand revenue held because of a specific re-weighting you chose. Then say which part you could not attribute. Honesty about the unattributable part reads as competence, not weakness.

What belongs on an SEO manager resume, and what should come off?

On: the shape of every property you ran (URL count, stack, locales, team size, developer capacity), outcomes with their baseline and measurement named, migrations and the redirect and monitoring work around them, the reporting you built and who uses it, engineering throughput you improved, and the AI-visibility work you can evidence (crawler access fixes, structured data, prompt-panel citation share, feed quality). Off: certificate lists, tool logos, self-rated skill bars, "passionate about SEO", rank screenshots, unanchored percentages, and any claim you could not reconstruct under questioning. Include the tools, but in the context of the work rather than as a cloud, because the screener searches for the names and the practitioner reads for the decisions.

What do SEO manager interviews actually test?

Diagnosis under incomplete information, prioritisation, and whether you can get work shipped. The most common exercise is a site audit, either live for twenty minutes or as a take-home with a presentation, and what is being scored is whether you can reach three prioritised problems with impact and effort attached rather than forty findings from a crawler. Expect a symptom question (traffic fell, pages are not indexed, a migration broke something) where asking what moved, for which segment, and what shipped that week scores higher than naming a cause immediately. Expect an engineering conversation about rendering, tickets and the cost of your asks. Expect to be asked what you would not do in your first quarter, and to be asked how you would get cited in AI answers, where naming the real submission surfaces and saying plainly what is unknowable is the correct answer.

How much does an SEO manager earn, and where can I find a real number?

There is no authoritative SEO salary table, so use two sources together rather than any single quoted band. The US Bureau of Labor Statistics has no SEO occupation: specialist-level search work falls under OES code 13-1161 (Market Research Analysts and Marketing Specialists) and people-managing or budget-owning roles under 11-2021 (Marketing Managers), both published with national and metro percentiles annually and roughly a year of lag, and neither isolating search work. The better source for your exact level is live postings: pay-transparency laws in states including California, Colorado, New York, Washington and Illinois mean a large share of ads carry a real band the employer has to honour. Collect twenty for your title, level and market, and sort them by employer type, because agency, publisher, SaaS and enterprise platform pay differ substantially at the same level. Tag each ad for whether it asks for AI visibility scope and compare the two groups yourself rather than trusting a quoted premium.

How do I get an SEO job with no professional experience?

Three routes work. Agency junior or executive roles are the highest-volume entry point, pay least and teach most, because you see many sites in a year. An internal move is often easier than an external one if you already work in content, paid search, analytics, web development or ecommerce merchandising; volunteer for the organic work nobody owns and document it. Freelance for two or three small businesses, priced low and scoped in writing, to generate real before-and-after material. In all three cases, run an owned property alongside: a small site on a subject you know, with a public change log where each entry states a change, a date, a hypothesis and a measured outcome. Ten honest entries over six months is more persuasive than any certificate, because an interviewer can check it.

Put this on a resume in about a minute

Paste your history once and point it at the SEO Manager posting you are looking at. No account, no card.

Build my resume free More roles

Related roles