Why Perplexity, Claude, and Gemini cite different sources.

Perplexity, Claude, and Gemini cite different sources because they run on different search indexes, different ranking rules, and different training data. Ask the same question in all three and you get three sets of citations. This is not a bug. Each tool decides on its own what counts as a trustworthy source, and those decisions rarely line up.

For a business owner trying to get found, this matters more than it looks. There is no single AI search index to win. Getting cited by one tool tells you almost nothing about the other two. So before you spend a dollar chasing AI visibility, it helps to understand why these tools disagree — and what that means for where you put your effort.

Why do these three tools disagree in the first place?

They disagree because each one uses a different pipeline to find, rank, and quote sources. An AI answer engine does two jobs. First it retrieves documents from the web. Then it writes an answer using those documents. Both steps differ across tools, and both steps change which sources you see cited.

Perplexity runs its own web index and also pulls live results from search partners. It leans hard on retrieval — it wants fresh pages and quotes them directly. Its answers read like a research assistant that just opened ten tabs.

Claude, made by Anthropic, added web search later and uses it more carefully. Claude answers a lot from its training data first, then reaches for the web when the question needs current facts. When it does cite, it tends to prefer a smaller set of sources it treats as reliable.

Gemini sits inside Google's world. It draws on Google's index and on Google's understanding of a page's authority. If Google already trusts your site in regular search, Gemini is more likely to surface you than a tool that has never crawled you.

Three pipelines. Three sets of rules. Three answers.

What each tool actually looks for in a source

Each tool weighs freshness, authority, and clarity differently, and that weighting decides who gets quoted. Knowing the bias of each one tells you where your page has a shot.

Perplexity favors freshness and direct relevance. It wants a page that answers the exact question, published or updated recently, with the answer near the top. If your page buries the answer under three paragraphs of intro, Perplexity may skip it for a competitor who put the answer in the first line. This is the same discipline good pages have always needed — a clear question, a clear answer, no throat-clearing.

Claude favors sources that read as careful and well-structured. Because it quotes fewer sources, the bar to get quoted is higher. Clean structure, plain claims, and a page that does not contradict itself all help. Claude is less impressed by keyword stuffing and more by whether the page actually holds together.

Gemini favors what Google already favors. That means the whole machinery of search quality applies — site reputation, page experience, and whether the page satisfies the intent behind the query. If your site loads slowly, that hurts you in Google search, and it can hurt you in Gemini too. We wrote about the mechanics of this in why your slow site is a sales problem, not an IT problem — the same speed signals that cost you clicks can cost you citations.

The common thread across all three is clarity. A page that answers the question plainly, loads fast, and does not lie is the page most likely to get cited by more than one of them.

Why the same page ranks in one tool and vanishes in another

A page can win in Gemini and disappear in Perplexity because the two tools never looked at the same set of candidate pages. Retrieval happens before ranking. If a tool's index does not contain your page, no amount of ranking quality will save you — you were never in the running.

This is the part most people miss. They assume every tool sees the whole web. They do not. Perplexity's index and Google's index are different sizes, refreshed on different schedules, with different coverage of small business sites. A new page on a small law firm's site might get into Google's index within days and into Perplexity's much later, or the other way around.

Coverage also varies by topic. Ask about a fast-moving news story and Perplexity's freshness bias wins — it will have the newest pages. Ask a settled question about, say, the difference between a will and a trust, and Gemini's authority bias wins — it will surface the established, well-linked pages Google already trusts.

So the same page can look invisible in one tool and dominant in another, and neither result tells the full story. You have to check each tool on its own terms.

What this means for a small business trying to get cited

The practical answer is that you cannot optimize for three tools separately, so you optimize for the thing all three reward. That thing is a clear, fast, trustworthy page that answers a real question. Chasing the quirks of any single engine is a poor use of a small budget. Building a page that any engine would be glad to quote is not.

Here is the order we work through with clients.

First, make the answer easy to find on the page. Put the direct answer in the first sentence under each heading. AI engines pull the opening lines of a section as their quote. If your answer is buried, it does not get pulled. This is a writing habit, not a technical fix, and it costs nothing.

Second, make the page fast and stable. Speed and layout stability feed Google's quality signals, which feed Gemini. They also affect whether a page gets crawled deeply in the first place. The three numbers that matter here are covered in Core Web Vitals: the three numbers that decide if Google bothers. A slow page is a page fewer engines bother to read.

Third, earn real authority. Get cited by other sites in your field. Publish under a named author with a real background. Keep your business information consistent across the web. Gemini reads these signals directly through Google, and Claude reads them indirectly through the reputation baked into its training data.

Fourth, keep the page current. Update it when facts change. Add the date. Perplexity rewards freshness, and a page that was last touched three years ago looks stale to a freshness-hungry engine.

Do those four things and you are not optimizing for Perplexity or Claude or Gemini. You are building the kind of page all three tend to trust. That is the whole idea behind our Search Foundations work — get the base right, and visibility follows across engines instead of one at a time.

We saw this play out with McShanes Solicitors. The work was not chasing any single AI tool. It was fixing the foundation — clear service pages, honest answers to the questions clients actually ask, and a site fast enough to crawl. Once the foundation held, the firm started showing up across search surfaces, not just one.

Should you try to game one tool over the others?

No. Building for a single engine is a bet you will lose, because these tools change their pipelines constantly and none of them publishes the rules. A page tuned to exploit one tool's quirk today can drop out tomorrow when that tool updates its retrieval or ranking. The effort is high and the shelf life is short.

There is a deeper reason too. The tricks that might nudge one engine are the same tricks that erode trust with the others. Keyword stuffing, thin pages spun up at volume, fake authority — these get you flagged, not cited. Our differentiator has always been what we refuse to do, and gaming a single engine sits firmly on that list.

The durable move is boring on purpose. Write pages that deserve to be quoted. Answer the question in the first line. Load fast. Tell the truth. Keep it current. That page earns citations across engines and keeps earning them after the next model update, because it never depended on a loophole in the first place.

Where this breaks down

None of this guarantees a citation from any specific tool on any specific day. AI engines are not transparent, they change without notice, and two people asking the same question in the same tool can get different answers. We cannot promise you a spot in Perplexity's answer box next Tuesday, and anyone who does is selling you something.

What we can say is that the foundation work compounds. A clear, fast, trustworthy page is the only asset that pays off across every engine, this year and next. Chasing individual tools is a treadmill. Building the foundation is a floor you stand on.

Search first. Funnel aware. Foundation always. The engines will keep disagreeing with each other. Your job is to build the one page they all struggle to ignore.