AI search news ·
Perplexity cited 546 sources to ChatGPT's 252 on the same questions — and a tool that reads answer text sees 2% of it
There are two ways to find out what an AI engine cited. You can read the citation payload the engine returns alongside its answer, or you can search the answer's text for URLs. They sound equivalent. They are not, and the gap is not small. Across our production corpus, 523 of 523 ChatGPT answers carry a URL typed into the answer text — 100%. For Perplexity the figure is 11 of 525 — 2.1%. Meanwhile, on a separate panel where all four engines answered identical questions and we read the payload instead, Perplexity returned 546 citation entries to ChatGPT's 252. The engine that cites the most is the engine a text-scraping instrument can barely see. We are publishing the ratio because we used to be on the wrong side of it.
Primary source: AskedAbout — production audit corpus (cut 2026-07-29) and four-engine panel run (2026-07-28)
The two instruments, side by side
This is a follow-up to yesterday's piece, which established that the two methods differ. What is new here is the size of the difference, measured per engine — because that is the number that decides whether a tool you are paying for can see your Perplexity visibility at all.
Instrument A — read the citation payload. Twelve buyer questions, three samples each, four engines, run 2026-07-28. Every engine gets the same 36 answers' worth of opportunity:
| Engine | Answers returning citations | Citation entries | Distinct domains |
|---|---|---|---|
| Perplexity | 36 of 36 | 546 | 157 |
| Google Gemini | 34 of 36 | 445 | 280 |
| OpenAI (ChatGPT) | 36 of 36 | 252 | 144 |
| Anthropic (Claude) | 25 of 36 | 222 | 97 |
Instrument B — search the answer text for URLs. Our production audit database, re-cut this morning: 1,048 completed answers across 177 audits in four local verticals — dentists, home services, med spas, personal-injury law. This is the older instrument, and it reads exactly one engine:
- ChatGPT: 523 of 523 answers (100%) contain at least one in-text URL, yielding 1,105 distinct registrable domains.
- Perplexity: 11 of 525 answers (2.1%), yielding 12 distinct domains.
Put the two together and the conclusion is unavoidable. Perplexity is not citing less — on identical questions it cites 2.2× more than ChatGPT. It simply delivers those citations as structured metadata rather than typing them into its prose, which is a presentation choice, not a visibility one. A tool that only reads prose records that behaviour as an absence.
Why this matters if you are buying an AI-visibility tool
- 1A report showing “no Perplexity citations” may be describing the tool, not your business. Before you conclude you are invisible on Perplexity — or that Perplexity does not matter — ask the vendor which of the two instruments produced the number. It is a fair question with a short answer.
- 2The error is directional, and it flatters ChatGPT. Any cross-engine comparison built on text scraping will overstate ChatGPT's share of citations and understate Perplexity's, Gemini's and Claude's. If a dashboard tells you ChatGPT is where all your citations are, check whether it could have seen the others at all.
- 3Absence of a link in the prose is not absence from the answer. This is the same asymmetry running underneath the finding that 24.7% of ChatGPT's brand mentions carry no link whatsoever. Being named, being linked, and being cited in the payload are three different events, and most tools collapse them.
We are not reporting this from a position of having got it right. Our own local-vertical corpus still runs on Instrument B, which is why the Perplexity column in it is nearly empty and why we label it a ChatGPT cut every time we publish from it. The four-engine panel was moved to the payload instrument on 2026-07-28. Migrating the larger corpus is the next job, and we will publish the before-and-after when it is done rather than quietly restating the old numbers.
If you want a reading that does not depend on a vendor's instrument choice, run the free check — we ask the engines the questions your customers actually ask and show you the verbatim answers, so you can see for yourself what was named and what was cited.
Why does my AI visibility tool show zero Perplexity citations?
The most likely reason is that it extracts URLs from the answer text. Perplexity returns its sources as a structured citation payload and rarely writes links into the prose — in our corpus, 11 of 525 Perplexity answers contained an in-text URL, against 523 of 523 ChatGPT answers. On identical questions read via the payload, Perplexity returned 546 citation entries to ChatGPT's 252.
Which AI engine cites the most sources?
In our 2026-07-28 panel — twelve questions, three samples, four engines, 36 answers each — Perplexity returned the most citation entries at 546, followed by Gemini at 445, OpenAI at 252 and Anthropic at 222. Gemini returned the most distinct domains (280) despite fewer entries than Perplexity. Claude declined to cite anything in 11 of its 36 answers.
How can I tell which method a tool uses?
Ask whether the citation count comes from the engine's returned citation array or from parsing the answer text, and ask for the per-engine answer coverage — what share of answers on each engine yielded any citation at all. A tool reading text will show near-100% coverage on ChatGPT and near-zero on Perplexity. That signature is hard to hide.
See your number
A free 60-second check shows what AI says about you.
Running this for clients? The $249 agency 5-pack audits five businesses, white-labeled.
Who runs this
- Built and operated by Sensara LLC, Atlanta, Georgia — about us and how the audit works.
- See what the report looks like before you run anything — score per engine, the competitors AI names instead of you, and a fix plan.
- We run the same audit on ourselves every week and publish the result: in the latest run AI named AskedAbout in 1 of 144 answers. We report our own numbers the way we report yours.