Only 11% of domains cited by ChatGPT are also cited by Perplexity. That figure comes from The Digital Bloom’s analysis of more than 680 million AI citations, and it should change how you plan content.
Rank on Google and you get one outcome. Get quoted inside an AI answer and you get another. Related jobs, not the same job.
This guide covers how each engine finds and picks its sources, what all three reward equally, and the practical checklist to get cited by Perplexity, ChatGPT and Gemini without writing three versions of every page.
Key takeaways
- The three engines pull from different indexes, so a page can be cited by one and invisible to the other two.
- Perplexity runs a live retrieval on almost every query, which makes crawler access, page speed and freshness disproportionately important.
- ChatGPT leans on its own crawler plus index partnerships, and rewards brands with consistent mentions across third-party sites, not just their own.
- Gemini and AI Overviews start from Google’s index, so your existing SEO is load-bearing rather than obsolete.
- One well-structured page can satisfy all three. Three separately optimised versions is wasted budget.
- Perplexity is currently the only one of the three that reliably sends measurable referral traffic into GA4.
Do Perplexity, ChatGPT and Gemini use the same sources?
No. The three engines retrieve from different indexes and weight different signals, so overlap between their cited sources is low. Perplexity searches the live web on each query. ChatGPT combines its own crawler with index partnerships. Gemini works from Google’s index and ranking signals. Same question, three different source sets.
That matters commercially for one reason. If you only track your visibility in ChatGPT because it is the one your buyers mention, you are measuring roughly a tenth of the picture.
It also explains a pattern I see constantly on audits: a client ranks well, appears in Google’s AI Overview, and is completely absent from Perplexity. Nothing is broken. The page was never retrievable by a live crawler in the first place.
How does Perplexity choose which sources to cite?
Perplexity searches the web in real time, retrieves a set of pages, extracts the most relevant passages, and cites the handful that best answer the question. It documents two separate systems: a crawler that indexes sites, and a user-agent that fetches pages live on behalf of someone asking a question right now.
The practical consequence is that access is the whole ballgame.
If your firewall, CDN or security plugin blocks either agent, you are not in the running. Neither is a page that takes four seconds to load, because a live fetch will not wait around.
Freshness carries real weight too. A page dated eighteen months ago competes badly against one updated last month, even when the underlying advice has not changed. Reviewing and re-dating your best commercial pages quarterly is one of the cheapest wins available.
Perplexity also shows its sources openly, which makes it the easiest engine to audit. Ask it a question your buyers ask, and you can see exactly who is being quoted instead of you.
How does ChatGPT decide what to cite?
ChatGPT answers from a mix of trained knowledge and live retrieval, and only shows citations when it has searched. Its retrieval draws on its own crawler and index partnerships rather than Google’s index, so Google rankings do not transfer automatically. Third-party mentions carry unusual weight here.
That last point is the one most UK SMEs underestimate.
When ChatGPT recommends suppliers, it frequently leans on directories, listicles, review sites and trade press rather than the supplier’s own website. Being written about elsewhere is a ranking signal in its own right. Our breakdown of how ChatGPT chooses which businesses to recommend goes deeper on the mechanics.
Two housekeeping items pay for themselves. Check that OpenAI’s crawlers are permitted in your robots.txt, using OpenAI’s published bot documentation rather than guesswork. And check that your business details are identical everywhere they appear online. Inconsistent names, addresses and service descriptions make it harder for any model to associate a citation with you.
How does Gemini source its answers?
Gemini and Google’s AI features are grounded in Google Search. They start from Google’s index, its ranking signals and years of entity data, then break a query into related sub-questions and pull sources for each. Traditional SEO is the entry ticket, not a separate discipline.
Google’s own position is that there is no special AI ranking system to game. Its guidance on AI features points back at standard fundamentals: crawlable pages, useful content, clean structure.
That is genuinely good news for anyone who has invested in SEO. It is bad news for anyone hoping GEO is a shortcut around it.
The wrinkle is that ranking first does not guarantee being the cited passage. Gemini selects the passage that most cleanly answers the sub-question it is working on. Deep, well-signposted sections beat one long undifferentiated page.
Perplexity vs ChatGPT vs Gemini at a glance
| Perplexity | ChatGPT | Gemini | |
|---|---|---|---|
| Primary source of answers | Live web retrieval on each query | Own crawler plus index partnerships, and trained knowledge | Google’s index and ranking signals |
| Citations shown | Consistently, inline | Only when it searches | Variable, not every claim |
| Weights most heavily | Crawl access, freshness, factual density | Third-party mentions, entity consistency | Existing search performance, structure |
| First thing to fix | Crawler access and page speed | Presence on directories and trade press | Technical SEO and passage clarity |
| Trackable referrals | Yes, most reliably of the three | Partly, via its citation tagging | Rarely |
What do all three engines actually have in common?
All three reward the same underlying writing: self-contained passages that answer one clear question without needing surrounding context. Question-led headings, a direct answer in the first sentences, tables for comparisons, and specific verifiable facts. The retrieval differs. The extraction preferences barely do.
This is why one page can serve all three.
A model does not quote your page. It quotes a passage from it. If that passage says “this approach works well” it is unusable. If it says “Perplexity fetches pages live, so anything slower than three seconds risks being skipped” it can be lifted straight into an answer.
We have set out the specific formats that survive extraction in five content structures that get picked up by generative AI engines.
The checklist to get cited by Perplexity, ChatGPT and Gemini
Work through these in order. Most take under an hour.
- Unblock the crawlers. Check robots.txt, your CDN and any security plugin for PerplexityBot, OAI-SearchBot, GPTBot and Google’s crawlers.
- Fix load speed on your money pages. Under three seconds, ideally under two.
- Rewrite your H2s as the questions buyers ask. Not “Our Approach” but “How much does commercial cleaning cost in London?”
- Answer in the first 50 words under each heading. Fully, without preamble.
- Make paragraphs self-contained. No “as mentioned above”, no floating pronouns.
- Add one comparison table per commercial page. Models parse tables cleanly and quote them accurately.
- Date your pages and update them quarterly. Freshness is a live signal, particularly for Perplexity.
- Standardise your entity details. Identical business name, address, phone and service wording across your site, Companies House, your Google Business Profile and every directory.
- Build third-party mentions. Trade press, directories, partner sites, industry roundups.
- Baseline your visibility. Ask each engine ten real buyer questions and record whether you appear.
Sector context changes the priority order. For firms selling into facilities management, for example, buyer behaviour skews differently again, which we cover in GEO for facilities management companies.
How do you measure whether any of this is working?
Track two things: citations and referrals. For citations, re-run your ten baseline questions monthly and log where you appear. For referrals, build an AI traffic channel in GA4 filtering on the AI platform domains. Perplexity produces the most reliable referral data, and ChatGPT tags its citation links, so both are measurable.
Set the expectation early with whoever holds the budget. AI citation visibility moves in weeks, not days, and it compounds. Treat month one as a baseline rather than a result.
The commercial argument is straightforward. A citation reaches a buyer at the exact moment they are choosing a shortlist, before they have searched for anyone by name. That is earlier in the funnel than most of your paid activity reaches, and it does not carry a cost per click.
Frequently asked questions
Do I need a different content strategy for each AI engine?
No. You need one well-structured content strategy plus three different technical and off-site checks. The writing that gets extracted is near-identical across all three engines: clear question-led headings, direct answers, self-contained paragraphs. What differs is access and authority. Perplexity needs crawl permission and speed, ChatGPT needs third-party mentions, Gemini needs solid conventional SEO.
Which AI engine matters most for UK searches?
Gemini and Google’s AI features reach the largest UK audience by a wide margin, simply because Google remains the default search destination here. ChatGPT matters most for supplier shortlisting and recommendation queries. Perplexity has a smaller audience but a professional, research-led one, and it sends the most trackable referral traffic. For most UK SMEs, prioritise Google, then ChatGPT.
Will getting cited actually send traffic to my website?
Sometimes, and less than ranking would. Many AI answers resolve the query without a click. The value is often in the mention itself: appearing on a buyer’s shortlist before they visit anyone’s site. Perplexity is the exception and does drive measurable referrals. Judge the channel on enquiry quality and mention share, not sessions alone.
How long does it take to start appearing in AI answers?
Expect four to twelve weeks for structural changes to show up, depending on how often each engine recrawls your site. Perplexity can pick up changes fastest because it fetches live. Gemini follows Google’s crawl cadence. ChatGPT is the slowest of the three, particularly where the gap is third-party mentions rather than on-site structure.
Does blocking AI crawlers protect my content?
It prevents citation entirely. Blocking is a legitimate choice if your commercial model depends on gated content or ad-supported page views. For service businesses that need to be found and recommended, blocking removes you from consideration in exactly the place buyers are now looking. Decide deliberately rather than inheriting a default from a plugin.
Do I need an llms.txt file?
Not yet. It is a proposed standard with no confirmed adoption by the major engines, so treat it as a low-cost experiment rather than a priority. Fix crawler access, page speed and passage structure first. Those are the things all three engines demonstrably use today.
Get found in the answers your buyers are reading
Being cited across all three engines is a structural problem before it is a content problem. Most sites need a handful of technical fixes and a rewrite of how their key pages are organised, not a new blog schedule.
If you want that mapped out properly, our AI search optimisation services cover the full picture across GEO and AEO.
