Opinion
Generative Engine Optimisation: How to Get Your Business Cited by AI Search in 2026
July 8, 2026
Generative Engine Optimisation (GEO) is the work of getting your business named and cited inside the answers that AI tools write, rather than just ranked in a list of blue links. It matters now because the answer is increasingly the whole result. Google's AI Overviews already appear on roughly half of searches and reach about two billion people a month, and when one shows up, the organic click-through rate on that query falls by more than 60% (Seer Interactive, 2026). If the AI writes the answer and your competitor is the one it quotes, the click was never on offer in the first place.
We've spent the past few months doing GEO on our own site, not as a theory but as a live project across our service and industry pages. This is what actually moved the needle, what didn't, and how you can apply the same thinking without burning a quarter on it.
Classic SEO earns a position in a ranked list. The user still clicks, still lands on your page, still reads your words in your layout. GEO earns a mention inside a synthesised answer that the user may never click away from. The unit of success changes from "rank" to "citation," and the audience changes from a person scanning results to a model deciding which sources to trust and paraphrase.
That shift has three practical consequences. First, being technically indexable is table stakes, not an achievement. Second, the model reads your content for extractable claims, not for keyword density. Third, your brand needs to appear across the wider web, not just on your own domain, because models weigh how often and how consistently you're mentioned elsewhere.
We started calling our top metric Share of Model: out of all the times an AI tool answers a question in our space, how often does Foresight Mobile get named? It's a harder number to pin down than a keyword ranking, and it's the one that now maps to pipeline.
The behaviour has already shifted. ChatGPT reports more than 800 million weekly users, Perplexity handles around 780 million queries a month, and Google's AI Mode passed 75 million users by December 2025 with generative revenue up almost 400% year on year (Alphabet Q4 2025). These aren't early adopters kicking the tyres. They're people getting answers to commercial questions ("best app development agency in Manchester", "how much does an MVP cost", "fractional CTO vs full-time") without visiting a single website.
There's a genuine upside for whoever gets cited. Brands that appear in AI Overviews earn roughly 120% more organic clicks per impression than uncited brands on the same query. The answer box isn't only taking clicks. It's redistributing them toward the sources it trusts. GEO is the work of becoming one of those sources.
The biggest mistake we see is treating AI search as a single target. It isn't. In 2026 research, only 11% of domains cited by ChatGPT were also cited by Perplexity (Omnibound, 2026). Ranking well in one does almost nothing for the other.
The reason is plumbing. ChatGPT's live web results are powered by Bing, so if you're not indexed in Bing, you effectively don't exist for a huge slice of ChatGPT's cited answers. Wikipedia alone accounts for close to half of ChatGPT's top citations, which tells you how heavily it leans on established, encyclopedic authority. Perplexity casts a wider net across the live web and rewards fresh, well-structured pages. Google's AI Overviews draw on its own index and increasingly on its Preferred Sources feature, which Google expanded for AI Overviews in May 2026.
So the honest version of GEO isn't "optimise for AI." It's: get indexed in Bing, earn the kind of third-party mentions that feed authority-led engines, and publish the fresh, structured, quotable content that the live-web engines reach for. Different levers, same content foundation.
When we audited what was working on our own pages against what the research predicts, the same handful of things kept coming up. Researchers at Princeton and Georgia Tech tested this directly in their GEO study and found that adding citations, statistics, and quotations to a page can lift its visibility in AI-generated answers by up to 40%. That matched our experience closely.
Models lift the sentence that most cleanly answers the query. If your page buries the answer under three paragraphs of throat-clearing, it doesn't get extracted. We rewrote headings as real questions ("Why does a construction business need a mobile app?") and put a direct, standalone answer immediately underneath. Not a tease. The actual answer, phrased so it stands on its own if a model quotes just that line.
Vague copy ("apps can improve efficiency") gives a model nothing to cite. Specific, sourced claims ("maintenance typically runs 15% to 25% of the original build cost per year") give it something worth quoting. We went through our service pages and replaced soft assertions with dated, attributed statistics. Where we didn't have a source, we cut the claim rather than invent one, because a wrong stat is worse than no stat when a model repeats it verbatim.
The GEO study found direct quotations were one of the strongest single levers. We added named quotes from the people who actually did the work, tied to real project detail. This is the same E-E-A-T signal Google has rewarded for years, and generative engines lean on it even harder because a named human with specifics is easier to trust than anonymous marketing prose.
This is the one most teams skip because it's the hardest. Ahrefs' analysis of AI Overviews across tens of thousands of brands found branded web mentions to be the single strongest correlate with getting cited. Not backlinks. Mentions. Your name appearing in directories, roundups, guest posts, podcasts, and industry reports, next to the topics you want to be known for. We treat this as our number-one priority now, above almost any on-page tweak.
Structured data won't force a citation, but it removes ambiguity about who you are, what you sell, and how your content is organised. We keep JSON-LD tight and accurate (Organization, Service, FAQPage where FAQs exist) and validate every change against both Google's Rich Results Test and the Schema.org validator before it goes live. Broken schema is worse than none.
Not everything the AI-search crowd recommends holds up. We tested the idea of serving a plain-text .md copy of every page and an llms.txt file, on the theory that models prefer a clean machine-readable version. The evidence says don't bother. Ahrefs looked at 137,000 domains in May 2026 and found 97% of llms.txt files received zero requests. Other server-log studies from the same period found effectively no AI-bot fetches of per-page .md files. We kept a single root llms.txt because it's cheap, and dropped the rest. It was time we got back.
The broader lesson: AI search is new enough that a lot of confident advice is untested. Prefer tactics with either controlled study backing or your own measured results, and treat everything else as a hypothesis.
Generative engines strongly favour recent content, and there's evidence of a citation cliff where pages that haven't been updated in roughly 90 days start losing their place in answers. For a topic that moves as fast as mobile (OS releases, framework updates, App Store policy changes), this rewards a publishing cadence more than a one-off content sprint. We now aim to publish something on genuinely fresh developments quickly, because being the first credible source on breaking news is one of the few reliable ways onto a Google Top Stories carousel and into AI answers about it.
Rankings won't tell you. We track three things instead. First, Share of Model: run your target questions through ChatGPT, Perplexity, Gemini, and Claude on a schedule and count how often you're named. It's manual and imperfect, and it's still the truest signal. Second, Google Search Console's generative AI performance reports, which Google began exposing in June 2026, so you can see impressions and clicks originating from AI surfaces. Third, referral traffic from AI domains in your analytics, which is small today but growing and worth a labelled segment.
Give it time. AI citation patterns lag content changes by months, not days, because models and their retrieval indexes refresh on their own cadence. We plan on a three-to-six-month horizon before judging whether a change worked.
If you're starting from scratch, do it in this order. Confirm you're indexed in Bing, because it gates a large share of ChatGPT citations. Rewrite your most commercial pages to answer their core question in the first sentence. Add sourced statistics and named quotes to those same pages. Get your schema accurate and validated. Then, and this is the long game, start earning brand mentions off your own site through directories, guest content, and PR. On-page work is a few weeks. Off-page authority is the quarter-long project, and it's the one that compounds.
None of this is exotic. It's disciplined content and PR aimed at a reader who happens to be a language model as much as a person. The teams that win at GEO in 2026 aren't gaming anything. They're just being the clearest, best-sourced, most-mentioned answer to the question their customers are actually asking.
GEO is the practice of structuring and promoting your content so AI tools like ChatGPT, Perplexity, and Google AI Overviews name and cite your business in the answers they generate. Where SEO aims for a ranked link, GEO aims to be the source the AI quotes.
It's an extension, not a replacement. GEO builds on the same foundations (crawlable content, accurate structured data, genuine authority) but optimises for being extracted into an answer rather than clicked from a list. Strong SEO fundamentals make GEO easier, and the two share most of the same work.
ChatGPT's live search runs on Bing, so being indexed in Bing is the prerequisite. Beyond that, it favours established, well-referenced sources, so accurate content, named expertise, and third-party mentions of your brand all help. Ranking in Google alone won't get you there.
Plan for three to six months. AI models and their retrieval indexes refresh on their own schedules, so citation patterns lag your content changes. On-page improvements can register within weeks, but the brand-mention work that drives most citations compounds over a quarter or more.
We build the kind of clear, well-structured, genuinely useful content and product experiences that both people and AI engines reach for, and we can help you do the same for your site and your app.
Get in Touch to talk through your GEO and content strategy, or see how we approach app development end to end.
