How to Rank in ChatGPT, Perplexity and Google AI Overviews: The 2027 Playbook

A plumber in Leeds asks ChatGPT which local firms handle emergency boiler repairs. A founder in Bengaluru asks Perplexity to compare three payroll tools. Neither of them sees ten blue links. They see a short answer with three or four sources attached, and if your business isn't one of those sources, you weren't in the race.

That is why "how to rank in ChatGPT" has become one of the most common questions we hear from clients. The scale is real: OpenAI said in February 2026 that ChatGPT had reached 900 million weekly active users. And on Google, a Pew Research Center study of 900 US adults found that about one in five searches in March 2025 already produced an AI summary.

Below: how each engine works, what the research says gets cited, and the steps we'd take on your site tomorrow.

How to rank in ChatGPT: the short answer

To rank in ChatGPT, Perplexity and Google AI Overviews, make sure their search crawlers can reach your site, publish pages that answer specific questions clearly with original facts, and get your brand mentioned on trusted third-party sites. AI engines retrieve pages from a search index, then quote the clearest, most corroborated sources.

That's the snippet version. In practice, "ranking" in AI search is two problems. Retrieval: does the engine find your page when it searches on the user's behalf? Selection: once your page is in the pile, is it worth quoting? Classic SEO mostly solves the first. The second is where generative engine optimization (GEO) earns its keep.

How each AI engine picks its sources

"AI search" isn't one system. Each engine has its own crawler, its own index (or a borrowed one) and its own habits. Here's what the platforms themselves have published.

Google AI Overviews and AI Mode

Google's AI features documentation is refreshingly direct. To be shown as a supporting link, "a page must be indexed and eligible to be shown in Google Search with a snippet." It also states there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."

The important detail is something Google calls query fan-out. Both AI Overviews and AI Mode may issue "multiple related searches across subtopics and data sources" to build one answer. So a single question like "best running shoes for flat feet" can trigger behind-the-scenes searches about arch support, cushioning, pronation and specific models. Your page can be cited for a sub-question you never targeted.

ChatGPT search

ChatGPT's web answers depend on OpenAI's search crawler, OAI-SearchBot. OpenAI's publisher FAQ says "any public website can appear in ChatGPT search" and the main instruction is simple: don't block OAI-SearchBot. It also notes that OpenAI may surface links it finds through third-party search providers, which is why your visibility in traditional search engines still matters for ChatGPT.

One handy detail: ChatGPT adds utm_source=chatgpt.com to referral links, so you can see that traffic in GA4 without guesswork.

Perplexity

Perplexity cites heavily and visibly, usually with numbered sources above or beside the answer. Its crawler documentation says PerplexityBot is "designed to surface and link websites in search results on Perplexity" and "is not used to crawl content for AI foundation models." So blocking it does nothing for training and everything to hide you from Perplexity's answers.

Microsoft Copilot and Bing

Copilot is grounded in Bing's index, and Microsoft now shows you some of the data. In February 2026 it launched an AI Performance report in Bing Webmaster Tools (in beta) that counts citations across Copilot, Bing's AI summaries and select partners, along with sample "grounding queries" the AI used to find your pages. No other platform offers anything as close to an AI citation console.

What the research says gets cited

Most AI search advice online is opinion dressed up as fact. Here's what the larger independent datasets show, with their dates, because this area moves quickly.

Rankings still matter, but less directly than before

When Ahrefs repeated its AI Overview analysis across 863,000 keywords in March 2026, the share of citations from top-10 pages had halved to 38%, with the rest split almost evenly between positions 11 to 100 and pages outside the top 100. The researchers point to query fan-out as the likely reason. YouTube was the single most-cited domain.

ChatGPT shows an even weaker link to Google rankings. Semrush's 2025 study reported that pages ChatGPT cites rank in position 21 or lower for related queries almost 90% of the time. Our read: strong SEO gets you into the candidate pool, but answering the exact sub-question cleanly is what gets you picked.

AI engines lean on encyclopedias and communities

Profound analysed 680 million citations between August 2024 and June 2025. Its breakdown showed Wikipedia as ChatGPT's top source (7.8% of all citations), while Reddit led for both Perplexity (6.6%) and Google AI Overviews (2.2%). You can't control Wikipedia, and you shouldn't astroturf Reddit. But you can make sure your brand comes up when people discuss your category in public (more in Step 3).

Specific facts beat keywords

The most cited academic work on this is the Princeton-led GEO paper presented at KDD 2024. Testing different rewrites on a benchmark of queries, the authors found that adding statistics improved a source's visibility in generated answers by roughly 41%, adding quotations by about 40% and citing sources by about 27%. Keyword stuffing made things worse. It was a lab benchmark, not a live engine, so treat the exact percentages loosely. The direction matches what we see in practice.

Step 1: Let the right crawlers in

This is the least glamorous step and the one we'd do first, because it's the most common silent failure we find in audits. A developer copies a "block all AI bots" robots.txt snippet from a forum, or a security plugin starts challenging unfamiliar user agents, and the site disappears from ChatGPT and Perplexity answers without anyone noticing.

The key is to separate search crawlers (which decide if you get cited) from training crawlers (which collect content to train future models). You can make a different decision for each.

The sources here are the platforms' own pages: OpenAI's crawler overview, Perplexity's bot guide linked above, and Google's list of common crawlers, which states plainly that Google-Extended "does not impact a site's inclusion in Google Search nor is it used as a ranking signal."

Should you block training crawlers at all? For most small and mid-sized businesses we'd leave GPTBot open, since being part of what a model "knows" about your category can help in answers that don't trigger a live search. Publishers with licensed content are a different case.

Step 2: Write pages AI can quote

When an AI engine retrieves your page, it's looking for a passage that answers the sub-question cleanly. Long throat-clearing intros, vague claims and key facts hidden in images all work against you. The fix is mostly good editing.

1. Lead each section with the answerPut a 40 to 60 word direct answer right under each question-style heading, then expand. Say you run a dental clinic in Pune with three branches: a page titled "How much do clear aligners cost in Pune?" should state the price range in the first two sentences, then explain what changes it.
2. Use headings that mirror real questionsPull phrasing from People Also Ask boxes, sales calls and your support inbox. Fan-out queries are specific, so specific headings give the engine a clean match.
3. Add numbers you can stand behindOriginal data is the strongest citation magnet we know of. That can be as simple as "we installed 412 solar systems in Melbourne last year and the median payback was X years." If you cite someone else's number, link to it.
4. Show who wrote it and why they knowNamed authors with real credentials, first-hand photos and clear sourcing are trust signals for people and machines alike. Our guide to E-E-A-T and the trust signals Google rewards covers this in detail.
5. Make comparisons explicitAI answers love "X vs Y" structure. Tables with clear rows (price, best for, limitations) are easy to extract and hard to misquote.
6. Keep pages currentShow a visible "updated" date where freshness matters, and actually update the page. Stale prices and specs give engines a reason to cite someone else.

Step 3: Earn mentions off your site

If we could only pick one off-site signal for AI search, it would be unlinked brand mentions in relevant places. Ahrefs' study of 75,000 brands found branded web mentions had a 0.664 correlation with AI Overview visibility, branded anchor text 0.527, and raw backlink count just 0.218. The authors are careful to say correlation isn't causation, and so are we. Still, it fits how language models work: they repeat what the web repeatedly says.

Where mentions do the most good

• "Best of" and comparison articles in your niche. When someone asks ChatGPT for the best CRM for a 10-person agency, it often draws on exactly these listicles.
• Industry publications and podcasts. A guest spot on a well-known trade podcast creates show notes, transcripts and social posts that all name you.
• Community threads on Reddit, Quora and niche forums, where your team answers questions genuinely and with disclosure. No sock puppets. Moderators spot them, and so do users.
• Directories and review sites that matter in your sector: G2 and Capterra for software, Practo for Indian clinics, Trustpilot for UK retail, and your Google Business Profile for anything local.
• YouTube. Given its rise as the most-cited domain in AI Overviews in Ahrefs' 2026 data, a handful of clear, well-titled explainer videos is no longer optional for many categories.

Does this mean links are dead? No. Links still drive Google rankings, and Google rankings still feed the candidate pool for AI answers. What's changed is that a mention in the right article can be worth more than a link from a random high-authority site. A good link building and digital PR programme in 2027 should be measured on both.

Consistency matters too. If your company description, pricing or service area differs across your website, LinkedIn and directories, models will hedge or pick the wrong version. For local firms, this is the same discipline that drives local SEO.

Step 4: Technical and structured data basics

Technical SEO for AI search is mostly just technical SEO. A few points deserve extra attention.

• Render key content as HTML. Don't assume every AI crawler renders JavaScript the way Googlebot does. If your prices, specs or answers only appear after client-side rendering, test with JavaScript disabled and fix what vanishes.
• Keep snippet eligibility. A nosnippet or tight max-snippet rule also limits what Google's AI features can show. Use data-nosnippet on specific blocks rather than whole pages.
• Add accurate structured data. Organization, Product, LocalBusiness, Article and FAQ markup help engines connect facts to entities. Google's guidance is that markup must match visible content, and it is not a shortcut into AI answers on its own.
• Get indexed in Bing. Verify your site in Bing Webmaster Tools, submit your sitemap and consider IndexNow so changes reach Bing (and therefore Copilot) faster.
• Fix speed and mobile basics. Slow or broken pages are crawled less and convert worse when the AI visitor does arrive. Our technical SEO audit checklist walks through every check.

llms.txt: the honest verdict

llms.txt is a proposed standard, put forward by Jeremy Howard in 2024, for a Markdown file at the root of your site that gives language models a curated map of your most useful pages. It sounds sensible. The problem is that there's little evidence the AI search engines use it.

"FWIW no AI system currently uses llms.txt."John Mueller, Search Advocate, Google, on Bluesky in June 2025 (reported by Search Engine Roundtable)

The data since then hasn't changed the picture much. In June 2026, Ahrefs published an analysis of server logs from 137,000 domains. About 28% had an llms.txt file, yet 97% of those files received zero requests during the study period. Where files were requested, SEO audit tools and general crawlers made most of the requests. AI retrieval bots such as OAI-SearchBot and PerplexityBot accounted for about 1.1%.

Our position: if your developer can add one in 20 minutes, fine, especially if you have documentation that coding agents might read. Just don't let anyone sell it to you as an AI ranking factor, and don't spend budget on it that should go to content or mentions.

How to measure AI search visibility

Measurement is still messy. No platform gives you a full ChatGPT equivalent of Search Console, and anyone promising perfect "AI rank tracking" is overselling. Here's the stack we actually use.

Set expectations honestly with your team. Pew's data showed users clicked a link inside the AI summary itself in just 1% of visits. A lot of AI search value will show up as brand searches, direct visits and "I saw you recommended in ChatGPT" on sales calls. Add that question to your lead forms. It costs nothing and it's often the clearest signal you'll get. For the wider numbers, our roundup of SEO and AI search statistics collects the latest research in one place.

What's overrated (and what isn't)

A pattern has emerged in what gets hyped and what quietly works.

And one opinion that may be unpopular: if your traditional SEO is weak, don't start with GEO. Fix indexing, content quality and authority first through a solid SEO programme. The AI layer amplifies what's already there. It won't rescue a site Google doesn't trust.

What to do this week

You don't need a six-month roadmap to start. Here's a five-day plan.

1. Monday: audit crawler accessOpen yoursite.com/robots.txt and look for OAI-SearchBot, PerplexityBot, Bingbot or blanket disallow rules. Then check your firewall's bot settings.
2. Tuesday: run a 20-question testAsk ChatGPT, Perplexity, Copilot and Google the questions your buyers ask before they buy. Note who gets cited and which pages.
3. Wednesday: fix your top three pagesAdd a direct answer under each main heading, one original number and a named author with credentials.
4. Thursday: set up measurementCreate the GA4 AI-referral exploration, verify Bing Webmaster Tools and add "Where did you hear about us?" to your forms.
5. Friday: list ten mention targetsFind the comparison articles, podcasts and directories that AI engines cited in Tuesday's test and plan how to get included.