Why Your Business Doesn’t Show Up in ChatGPT — and How to Fix It

Short answer: To show up in ChatGPT, AI assistants cite pages that answer a specific question completely, in one place, with clear structure and verifiable specifics. They favour sources that are already well-linked and frequently mentioned across the web. To get cited, you need question-shaped pages with the answer in the first fifty words, real named specifics rather than generalities, valid schema markup, crawlable HTML that doesn’t require JavaScript to render, and mentions on third-party sites that AI models trust.

Something changed, and most businesses haven’t noticed yet

A client in Leads called us last spring because their traffic was down about 20% year over year while their rankings had barely moved. Same positions, fewer clicks. We looked at the queries: the informational ones the “what is,” “how do I,” “how much does” searches that fed the top of their funnel had collapsed. The transactional ones were fine.

Their customers hadn’t stopped searching. They’d stopped clicking, because they were getting the answer above the results.

That pattern is now well documented. Ahrefs‘ tracking found AI Overviews reduced clicks to top-ranking pages substantially their updated study put it around 58% on affected queries. Pew’s 2025 analysis of real browsing behaviour found users who saw an AI summary clicked a link roughly half as often as users who didn’t. Estimates vary wildly by study, methodology and query type, and 2026 data from Seer Interactive suggests some rebound. But nobody credible is arguing the effect is zero.

Here’s the part that makes it worth your attention rather than your panic: the traffic that does come through converts well. Reported conversion rates for AI-assistant referrals range from modestly better than non-branded organic — Diggity’s data showed about 31% higher — to several times better in some e-commerce datasets, while other studies found the opposite. The honest read is that the numbers are unsettled but the visits are qualified, because somebody who arrives after an AI has already answered their questions is much further down the decision than somebody who just clicked a blue link.

Fewer visitors, warmer visitors. That changes what you should be optimising for.

How AI assistants actually choose what to cite

Three different mechanisms, and they matter because they respond to different work.

Retrieval-augmented answers — ChatGPT with search, Perplexity, Google AI Mode run live searches, fetch a handful of pages and synthesise. These are winnable with content work in weeks, not years, because they’re reading the live web.

Training-data knowledge — the model answering from what it absorbed during training, with no live lookup. You influence this through the volume and consistency of how your brand is described across the web, and the timeline is long.

Grounding in structured sources — knowledge graphs, business profiles, review platforms, industry directories. This is why your Google Business Profile, your NAP consistency and your presence in credible directories matter more than they did in 2023.

Most people only work on the first one. The compounding happens in the second and third.

The Eight Things That Get You to Show Up in ChatGPT

Write pages shaped like questions. Not “Our Services” but “How long does medical credentialing take?” Then answer it in the first two sentences before you elaborate. AI systems extract answers; they can’t extract what’s buried in paragraph nine.

Could you lead with the answer, then earn the reading? Every page gets a forty-to-sixty-word direct answer near the top. This is the single highest-leverage change most sites can make, and it also improves human engagement, which is a nice thing about AEO; it rarely fights good writing.

Be specific enough to be useful. “Fast turnaround” is unciteable. “Most builds run six to eight weeks” is citeable. Numbers, timeframes, prices, named tools, named locations, named limitations. Vagueness is the enemy of citation, and it’s the number-one reason generic AI-written content never gets picked up: it has nothing in it worth quoting.

Cover the whole question on one page. AI systems prefer sources where the full answer lives in one place. Ten thin posts on adjacent slices of a topic lose to one thorough page. Consolidate your near-duplicates.

Mark it up. Article, FAQPage, LocalBusiness, Service, Organization, Product, BreadcrumbList — whatever fits, valid and tested in Google’s Rich Results tool. Schema doesn’t make you the answer, but it makes your content unambiguous to a machine deciding whether it understands your page well enough to quote it.

Make sure your content exists in the HTML. If your key copy only appears after JavaScript executes, some crawlers will never see it. Test with JavaScript disabled. This catches an uncomfortable number of expensive sites.

Get mentioned where models are looking. Industry publications, podcasts, review platforms, comparison sites, credible directories, genuine expert commentary. Brand mentions across trusted third-party sources influence what a model believes about you far more than anything you publish on your own domain. This is PR work wearing an SEO hat, and it’s the part agencies quietly skip because it’s hard.

Keep it current and date it visibly. Assistants prefer recent sources for anything time-sensitive. A page that says “updated February 2026” with figures that match will beat a stronger 2023 page on a question where the answer moves.

On llms.txt: it’s cheap to add and there’s little evidence yet that major assistants use it. Add it, don’t build a strategy on it.

What to stop doing

Stop publishing high-volume generic content. It has no citeable specifics, and it dilutes the authority of the pages that do. Stop scattering one topic across eight thin posts. Stop hiding your prices, timelines and process — those are exactly the questions people ask assistants, and refusing to answer them means someone else’s page answers them for you. And stop measuring only sessions, because if you’re not tracking referrals from chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com in GA4, your fastest-growing channel is currently invisible in your reporting.

How to measure any of this

Segment AI referrers in GA4 as their own channel group. Track branded search volume in Search Console — when AI mentions rise, branded searches usually rise with them, because people read the answer and then Google the company. Prompt the major assistants monthly with the real questions your buyers ask and log whether you appear and how you’re described. It’s manual and it’s the most useful hour of the month.

The uncomfortable truth

AEO isn’t a separate discipline you bolt onto SEO. It’s what SEO looks like when the intermediary got better at reading. Sites that were already fast, well-structured, genuinely expert and specific are getting cited. Sites that were winning on volume and keyword density are not.

If that sounds like bad news, it’s because it’s expensive. It’s also durable, which is more than could be said for most of what worked in 2019.

Frequently asked questions

How do I get my website to show up in ChatGPT?Publish pages that answer specific questions directly in the first fifty words, include verifiable specifics like prices and timelines, add valid schema markup, ensure your content renders in HTML without JavaScript, and build mentions on credible third-party sites.

Is AEO different from SEO?It overlaps heavily. AEO emphasises question-shaped content, direct answers, structured data and third-party brand mentions over keyword targeting and link volume. The technical foundations are the same.

How long does it take to get cited by AI search?Assistants that search live can pick up new pages within days to weeks. Influencing what a model knows from training data takes months and depends on mentions across the wider web.

Does schema markup help with AI search?It helps machines interpret your content unambiguously, which supports citation. It isn’t sufficient on its own — the content still has to answer the question better than the alternatives.

Should I add an llms.txt file?It costs almost nothing, so add it. Evidence that major assistants currently rely on it is thin, so don’t treat it as a strategy.

We build AEO foundations into every site we ship — structured data, answer-first content architecture and AI-referral tracking from day one. Every page is engineered to show up in ChatGPT and the assistants your buyers already use. Book a strategy call if you want to know where you currently stand.