Why your report links real sources for Perplexity — and shows none for ChatGPT, Claude, or Gemini

Open a PingMyBrand report and scroll to "Where AI looks," and you'll usually see one of two things: a grid of real domains with clickable links underneath them, or an honest empty state saying none of the tracked engines cited a source this scan. Look closer at which engine actually produced the links you do see, and a pattern holds almost every time: when there are citations, they trace back to Perplexity. ChatGPT, Claude, and Gemini can all name your brand — or a competitor's — in the exact same scan, and none of the three will ever hand you a link to click. That's not three broken engines. It's one structural fact about how these APIs answer a question today, and it's worth understanding exactly, because it changes what a blank citation list should and shouldn't tell you about your brand.

Two different requests, two different kinds of answer

Every engine in your report is answering the same 25 buyer questions, but the kind of request going out the door isn't the same for all four. ChatGPT, Claude, and Gemini are asked with a plain chat-completion call: send a prompt, get back a block of prose. Nothing about that request tells the model to go fetch a live web page and hand back the URL it used — and without that instruction, the response has no source-URL field to read, because there's nothing in it to extract. The model answers from what it already knows, and prose is all a plain chat-completion call was ever built to return.

Perplexity works differently by default. Its retrieval-first model is built to search the live web on essentially every question and hand back the actual pages it used, as a numbered citation list — not an optional add-on you have to explicitly request, but the default shape of the response itself. That's a real product difference between the two kinds of calls, not a matter of one engine trying harder than the other three.

Why a blank citation list doesn't mean you weren't cited

Here's the part worth getting exactly right, because it's easy to read an empty "Where AI looks" box as "we have no evidence anyone read your page" — and that's not what it means. ChatGPT, Claude, and Gemini can all retrieve and cite live pages; OpenAI, Anthropic, and Google each publish a way to turn on live web search for a given call, with real citation URLs coming back once it's enabled. PingMyBrand's calls to those three don't currently turn that mode on. So a blank citation list for ChatGPT, Claude, or Gemini on a given scan tells you exactly one thing: this scan didn't ask that engine to browse and report back a URL. It tells you nothing about whether the model, browsing on its own inside the real product a buyer actually opens, would find and cite your page.

Whether ChatGPT, Claude, or Gemini named you at all — the reading that actually drives your visibility score — is a completely separate, fully real measurement, taken from the same answer. The absence of a citation link sits next to that number. It isn't a comment on it, and it isn't evidence against you.

Why we haven't turned it on

Honest answer: it's a real, scoped change we've deliberately not shipped yet, not an oversight. Turning on live browsing for three more engines on every scan adds real per-call latency and real per-call cost on top of what a scan already spends — a decision that changes what a scan costs and how long it takes for every customer on the platform, not a quiet copy tweak. That's a genuine product tradeoff, the same category of open, named gap we wrote about when we fixed "Not measured" vs a real 0: we'd rather tell you plainly where the evidence stops than quietly imply three more engines are checking sources when they aren't. Grok — real infrastructure in our data model, not yet live in every scan — would face the exact same chat-completion-vs-retrieval split the day it joins the other four.

What to actually do with a blank "Where AI looks" section

Don't read it as a verdict on your brand specifically. Two things are still worth acting on, regardless of whether a link renders this scan:

• If Perplexity did cite sources this scan, that domain list is real, current evidence of who's winning the citation your buyer just read. Go pitch for inclusion on those pages, or match their shape — that read holds independent of what ChatGPT, Claude, or Gemini showed in the same run.
• If ChatGPT, Claude, or Gemini named you (or a competitor) with no citation attached, that's still the single most important line in the report. Being named without a link is a real, if less checkable, outcome — and it's the reading that actually moves your score, citation or not.

See the difference on your own brand

You don't have to take any of this on faith. Run the free scan, open your report, and put the "Where AI looks" section next to the per-engine breakdown directly above it. If Perplexity cited real domains and the other engines show none, that's the exact mechanical difference this post just walked through — not a bug in your report, and not proof of anything about your brand on its own. What actually matters is what's above it: whether each engine named you at all. Run the free scan — no signup, about a minute — and read that number for yourself, not a guess based on which boxes happen to have links in them.