Perplexity shows 12 sources.
Grok reads 44 pages and shows 4.

Method

Each question was written the way people type into a search box, for example "good dermatologist in bangalore for acne scars". We asked it once per engine, in a fresh chat, in each engine's normal web app.

For every answer we logged four things: the sources a reader could see, the brands or businesses it named, its length in words and, for Grok, the number of pages it said it had read.

Each engine shows sources its own way, so we counted what a reader actually sees:

• Perplexity:

  the number on its Sources panel.

• ChatGPT:

  its citation pills. A pill reading "RTINGS.com +1" counts as two.

• Grok:

  clickable links inside the answer, with its "pages read" figure logged separately.

Full results

The table below has one row per answer: 30 questions across 3 engines.

Columns

Scope and
limitations.

• One run per question. AI answers change from one run to the next, so each row is a snapshot.

• Two collection dates. Perplexity answered all 30 on 29 October. ChatGPT and Grok answered questions 1 to 10 that day and 11 to 30 on 31 October, after Grok's free plan paused us for high demand.

• Account plans. The ChatGPT account was on a paid Plus plan for the 31 October run.

• One location. Every search ran from India. Results elsewhere will differ.

• Different source displays. The three engines don't show sources the same way, so "sources shown" compares what a reader sees, not what each engine used.

• Small sample. 30 questions show patterns, not market-wide rates.

• Our interest. Rank in AI Overview helps brands improve their visibility in AI search, so we have a commercial interest in this topic.

How to
cite.