I spend most of my week looking at how recruitment and search firms show up in search, and the thing that has changed fastest over the past year is not rankings. It is what happens before the click. A hiring manager who used to type "engineering recruitment agency Birmingham" into a search box now often asks an AI assistant to name a few good ones, reads the answer, and only then starts opening websites. If your firm is not in that answer, the rest of your marketing never gets a look in.
Generative engine optimisation is the work of getting into those answers. This guide covers what it actually is, what the research shows genuinely improves AI visibility, what does not work despite what you may have been sold, and a practical order of tasks for a recruitment agency starting from scratch.
Generative engine optimisation, explained without the jargon
Generative engine optimisation, shortened to GEO and often written the American way as generative engine optimization, means shaping your content and your wider online presence so that AI tools find you, understand you and quote you when they answer a question.
The mechanics are worth understanding, because they explain everything that follows. When you ask an AI assistant a question, it does not recite something it memorised. It runs searches in the background, retrieves a set of pages, reads them, and writes an answer grounded in what it found. Google describes this as retrieval-augmented generation, paired with a technique it calls query fan-out, where a single question is broken into several searches behind the scenes.
That has two consequences for you. Your pages must be indexed and crawlable or they cannot be retrieved at all. And once retrieved, your content has to be easy to lift a sentence from, because the assistant is looking for something quotable, not something persuasive.
GEO vs SEO: what actually changes
People ask me whether GEO replaces SEO. It does not. It sits on top of it.
Everything technical stays the same. Crawlable pages, sensible site structure, fast loading, clean internal linking, pages eligible to appear in snippets. Skip any of that and you are invisible to AI tools for the same reason you are invisible in ordinary search.
What changes is the goal and the writing. In classic SEO you want a position, and a compelling title that earns a click. In GEO you want to be the source a machine draws from, which rewards a different style: direct answers near the top of the page, plain statements of fact, real numbers, named people, and links out to where your claims came from. The irony is that this is simply good, honest writing, which most agency websites do not do because they are busy selling.
Why recruitment agencies should care about AI search now
Two things are happening at once, and together they change the maths.
The clicks are going somewhere else
Zero-click searches, where someone gets their answer without visiting any website, reached 68% of Google searches in early 2026 according to analysis reported by Search Engine Land. When an AI Overview appears, organic click-through rate has been measured at 0.61%, against 1.62% when one does not appear. Business-to-business research queries have been hit particularly hard, losing around 22% of organic traffic to AI Overviews.
I want to be careful here, because this gets overstated. It does not mean SEO is finished. The people who still click through have read a summary and chosen to go deeper, so they arrive better informed and closer to enquiring. But it does mean the same ranking now delivers fewer visits than it did two years ago, and part of your visibility has moved into an answer box you do not control.
Your clients are asking AI tools to shortlist agencies for them
This is the part that matters most for recruitment. Gartner's research has found that six to ten stakeholders shape a typical business purchase, that buyers spend only around 17% of their time meeting suppliers at all, and that most now prefer to research without speaking to anyone. AI assistants fit that preference perfectly. They produce a shortlist in seconds, with no sales call attached.
So when a client asks which firms handle executive search in professional services, or which agencies specialise in warehouse staffing in the North West, an assistant names a few. There are somewhere around 28,000 to 30,000 recruitment agencies in the UK, turning over roughly £42.9 billion between them according to the REC, and nearly 80% of them employ fewer than ten people. Being one of three firms named in an answer is a very different position from being one of thirty thousand.
You will not see any of this in your analytics. There is no referrer, no click, no record. That is exactly why it is easy to ignore for another year.
What the research says actually works
This is where I want to be useful rather than speculative, because there is a lot of confident nonsense being sold under the GEO label.
The foundational study was published by researchers from Princeton and IIT Delhi and presented at the ACM SIGKDD conference. They tested nine ways of rewriting web content and measured how visible each version became in AI-generated answers, across thousands of queries. The findings are unusually practical.
The three techniques that moved the needle
Three changes stood out, and all three are about evidence rather than style.
Adding quotations. Including relevant quotes from credible people scored highest of everything tested, lifting visibility to 27.8 on the study's main metric against a baseline of roughly 19 to 20.
Adding statistics. Replacing vague qualitative claims with actual figures scored 25.9.
Citing sources. Linking out to credible sources for your claims scored 24.9.
Taken together, the researchers reported those three methods producing relative improvements of 30% to 40% on their position-adjusted word count metric and 15% to 30% on subjective impression, with the best methods reaching 41% and 28% respectively.
For a recruitment agency this translates neatly. A page that says "we place candidates quickly" is invisible. A page that says "our average time to shortlist across 140 engineering mandates last year was eleven days", quotes a named client on what that meant for them, and links to the salary survey it draws on, is the kind of page an AI tool can actually use.
Keyword stuffing made things worse
The study also tested keyword stuffing, the old trick of packing a page with repeated search terms. It scored 17.8, below the unoptimised baseline. Adding unique words scored 20.7, barely different from doing nothing. Purely stylistic edits, such as simplifying the language or making the tone more authoritative, produced small and inconsistent gains.
The lesson is that the tactics people reach for first are the ones that do not work. Rewriting your service page to say "recruitment agency" fourteen times will make it perform worse, not better.
The finding that matters most for smaller firms
Here is the result I find genuinely encouraging, and it rarely gets mentioned.
The gains were not evenly spread. Websites already sitting at the top of the results generally lost visibility when optimised, by around 20% to 30%. Websites ranking lowest gained enormously, roughly doubling their visibility, with the citation method producing a gain of about 115% for the lowest-ranked sites.
In plain terms, generative engine optimisation disproportionately helps the firm that is currently fifth rather than first. If you are a specialist agency who has never outranked the large generalist job boards, this is the first mechanism in years that works in your favour rather than theirs.
What Google says about optimising for AI search
Google published its own guidance on this, and it is refreshingly blunt about the things you do not need to do. I would read it before paying anyone for GEO work.
Its recommendations for what to do are familiar: meet the technical requirements so your pages are indexed and eligible for snippets, keep content crawlable, follow sensible JavaScript practices, provide a good page experience, and create what it calls "valuable, non-commodity content" that carries a genuine point of view rather than repackaging what already exists.
What does not work
The guidance is clearer still on what to ignore. According to Google you do not need to create machine-readable files, AI text files, special markup or Markdown versions of your pages. The llms.txt file that a lot of agencies are currently selling does not help you rank. You do not need to break your content into small chunks. You do not need to write in a particular style for AI. Structured data is not required for generative AI features, though it remains worth having for other reasons. And chasing artificial brand mentions does not work.
If someone pitches you a GEO retainer whose main deliverables are an llms.txt file and some schema markup, they are selling you things the search engine has publicly said are not required.
Where AI tools get their information about your agency
One more piece of context before the checklist, and it is the one that most changes how you should spend your budget.
Several industry analyses through 2025 and 2026 have found that the large majority of brand citations in AI answers come from third-party sites rather than the brand's own website, with estimates commonly landing between 82% and 89%. These come from vendors with an interest in the conclusion, so I would treat the exact figures with caution, but the direction is consistent and it matches how the tools behave. ChatGPT, for example, cites sources in the overwhelming majority of its responses and typically draws on around five of them per answer.
What that means practically is that your own website is necessary but not sufficient. Being named in industry publications, sector round-ups, directories, podcasts, professional bodies and client case studies published elsewhere carries real weight, because those are the pages the assistant retrieves when someone asks who the good firms are. Getting cited is a public relations problem as much as a technical one.
A generative engine optimisation checklist for recruitment agencies
Here is the order I would work through it. The early items are cheap and quick, the later ones compound.
Fix the technical foundations first
Nothing else matters if your pages cannot be retrieved. Confirm your key pages are indexed. Check that your service and sector pages are not blocked, orphaned or buried four clicks deep. Make sure content rendered by JavaScript is visible to crawlers, which catches out a surprising number of recruitment sites built around job board integrations. If you are unsure where you stand, a straightforward technical audit will tell you in a day.
Write pages that answer the questions clients actually ask
Most recruitment agency websites are organised around what the firm wants to say. AI visibility rewards sites organised around what clients ask.
Write down the real questions: what a retained search costs, how long a search takes, what happens if a placement leaves, how your fees compare to contingency, which sectors you cover and how deeply. Then give each one a proper answer in plain English, near the top of the relevant page, in a sentence that stands on its own if lifted out of context. That last point is the whole discipline in one line.
Put numbers, quotes and sources into your content
This is the highest-return change available to you, and it costs nothing but honesty.
Go through your main pages and replace every vague claim with a specific one. Fill rates, time to shortlist, retention at twelve months, number of mandates in a sector, average salary bands you recruit in. Add a named quote from a client or a consultant. Link out to the sources you rely on, whether that is REC data, an industry salary survey or official labour market statistics.
Three notes of caution. The numbers must be real, because a client will check them. Ask clients before quoting them. And do not name other firms or invent examples to fill space, since a fabricated figure that gets quoted back to you by an AI tool is considerably worse than no figure at all.
Build mentions on the sites AI tools already trust
Given how much of AI citation comes from third-party sources, this deserves genuine budget. Contribute to trade publications your clients read. Get listed accurately in the directories and professional bodies relevant to your sectors. Publish client stories that the client is willing to share on their own site too. Appear on sector podcasts. Support original data, such as a salary benchmark for your niche, since original figures get quoted far more often than opinion does.
Get your entity details consistent everywhere
AI tools have to work out that the various mentions of your firm refer to the same organisation. Make that easy. Use the same legal name, address and description everywhere. Keep your business profile accurate, particularly if you have several offices. Keep your team's professional profiles current, since people are part of how a firm is understood. Structured data is not required, as Google says, but it is a cheap way of stating plainly who you are and what you do.
Favour specific pages over general ones
A page about recruitment is competing with everyone. A page about interim finance recruitment for private equity backed businesses in the Midlands is competing with almost nobody, and it matches the shape of the questions people actually ask assistants. This is the same argument for niching that applies to the rest of your marketing, and AI search sharpens it considerably.
How to measure AI visibility
You cannot manage this without some form of measurement, and the honest answer is that the tooling is still immature. What I would do:
Build a list of ten to twenty questions a client would genuinely ask before hiring a firm like yours. Put them to the main assistants once a month and record which agencies get named, which pages get cited, and whether you appear at all. It is manual and it takes an hour, but it is the most direct measure available and it tells you more than any dashboard.
Alongside that, use Search Console's reporting on how your site performs in Google's AI features. Then watch the indirect signals: branded search volume, direct traffic, and enquiries where the prospect says they were already familiar with you and cannot quite say why. That last one is usually the first sign this is working.
Mistakes I see recruitment firms make
Treating it as a separate project. GEO is not a campaign that runs for a quarter. It is a change to how you write and where you appear.
Buying the technical trinkets. The files and markup being sold hardest are the ones Google has said are unnecessary.
Publishing more of the same content. Volume was a weak strategy in SEO and it is a worse one here. Ahrefs found that only 1.74% of newly published pages reach the top ten within a year, and that 72.9% of pages currently in the top ten are more than three years old. Improving ten existing pages beats publishing ten new ones.
Abandoning SEO. AI tools retrieve from search indexes. Your search performance is the input.
Leaving the numbers out. Almost every recruitment website I audit is full of claims and empty of evidence, which is precisely the content AI tools cannot use.
How long generative engine optimisation takes
Content changes work relatively quickly. Adding figures, quotes and sources to an existing page can change whether it gets cited within weeks, because AI tools revisit pages often and there is no ranking queue to wait in.
The other half is slower. Earned mentions, sector coverage and original data take months to accumulate, exactly as link building always did. I would expect to see movement in citations within one to two months of rewriting your main pages, and a meaningful change in whether you get named in shortlists somewhere between six and twelve months, depending on how visible your firm is to begin with.
Where to start if you only have a day
If you take one afternoon, do this. Pick your three most commercially important pages, usually your main service page, your strongest sector page and your about page. On each one, add a direct answer to the obvious question in the first hundred words, replace three vague claims with real numbers, add one named client quote, and link out to two credible sources. Then write your ten client questions down and put them to an AI assistant so you know where you stand today.
That is a genuine start, and it costs nothing.
Where we fit
We are an SEO agency for recruitment firms, and generative engine optimisation is now part of how we approach that work rather than a separate product we sell alongside it. In practice it means the technical foundations, the content rewriting described above, and the earned coverage that gets a firm named in the first place. If you want to talk about where your firm currently sits in AI answers, get in touch and we will run the questions your clients ask and show you what comes back.
If you would rather work through it yourself first, our guides on how to build a recruitment website and choosing a B2B SEO agency cover the groundwork this sits on top of.
Sources
- GEO: Generative Engine Optimization, Aggarwal et al., ACM SIGKDD, on which content changes improve AI visibility
- Google Search Central: AI features and your website, official guidance on what does and does not help
- Google zero-click searches reach 68% in early 2026, Search Engine Land
- Gartner sales survey on rep-free buying
- Recruitment Industry Status Report, REC, on UK industry size
- How long does it take to rank in Google, Ahrefs
