What AI reads before it recommends AI Customer Support tools

One question, put to the two AI engines that hand back the pages they opened.

The question, as it was asked

“What are the best AI customer support agents that automatically resolve tickets and deflect inquiries?”

asked Nov 4, 2026 · Perplexity and Google AI Overviews · US English

These are not the leaderboard's engines. The leaderboard and the debate ask ChatGPT, Claude and Gemini. This page asks Perplexity and Google, because they are the two that disclose what they read. What an engine reads here does not explain a rank there.

Who they named

The companies each engine's answer named, in the order it first mentioned them. Order of mention is not a rank — neither engine ranked anything.

Google AI Overviewsnot measured · asked 3 times

We could not read Google AI Overviews in this check

Asked 3 times and ended without a usable result. This engine is absent from every count on this page — it is not a zero.

What they read

Every page each engine retrieved while answering. Retrieved is not the same as relied on — the engines report the pages they pulled back, and no engine tells anyone which of them changed its answer.

PerplexityPerplexity retrieved 50 pages from 28 sites· 8 of those pages sit on a company's own site

Most-read sites

g2.comreview site11 pages
learn.g2.comreview site5 pages
usefini.comown site4 pages
sobot.io3 pages
irisagent.com2 pages
lorikeetcx.aiown site2 pages
resources.rework.com2 pages
1mind.com1 page
apnews.com1 page
automationanywhere.com1 page

Every page retrieved

Google AI Overviewsnot measured · asked 3 times

We could not read Google AI Overviews in this check

Asked 3 times and ended without a usable result. This engine is absent from every count on this page — it is not a zero.

The regulars

Sites this engine keeps coming back to. Page lists change from one check to the next; the sites are much steadier, which is why this is the only thing here counted across checks.

Perplexity
  • lorikeetcx.ai5/5
  • helpdesk.com5/5
  • usefini.com4/5
  • sobot.io4/5
  • resources.rework.com4/5
  • automationanywhere.com4/5
  • decagon.ai4/5
  • g2.com3/5
Google AI Overviewsnot measured · asked 3 times

Google AI Overviews has no measured checks behind this one, so it has no regulars yet.

Google AI Overviews has no measured checks behind this one, so it has no regulars yet.

What these numbers rest onOne question · the dates · how often each engine was asked

This page is about one question, not about the category. It says: when a buyer asks this exact question, here is what each engine retrieved and who each one named. It does not say these are the pages AI reads about this category — measured, one question finds only about a third of the sites the engines keep returning to across a whole market.

  • Retrieved, never cited. The engines hand back the pages they pulled; none of them says which ones it leaned on.
  • Per engine, never pooled. The engines read mostly different webs, so a combined page count would describe a list neither of them produced. The only cross-engine facts here are that both engines read a site and that both engines named a company.
  • Counts, never rates. Every number here counts a named thing. There is no percentage anywhere on this page.
  • An engine we could not read is said so plainly and enters no count. It is never shown as a zero.

When an engine returns nothing — the call fails, Perplexity answers without searching, or Google shows no overview — it is asked again, up to 3 times in all. Whatever the last attempt gave is what is published, with the number of attempts beside it.

  • Perplexity — asked once on Nov 4, 2026
  • Google AI Overviews — asked 3 times on Nov 4, 2026

We do not open the retrieved pages. Everything here comes from what the engines themselves handed back. Both engines are asked from the United States, in English.

These are not the leaderboard's engines. The leaderboard and the debate ask ChatGPT, Claude and Gemini. This page asks Perplexity and Google, because they are the two that disclose what they read. What an engine reads here does not explain a rank there.

Methodology version 1.1.0Read the full methodology →

This check is one of a set. The same question is put to the same two engines in every tracked category, once a month.

What the engines read across every category →

Earlier checks

Earlier checks of this category, each on the archive page for its date, beside the leaderboard and the debate it shares that date with.

Oct 18, 2026 →Oct 11, 2026 →Oct 4, 2026 →Sep 27, 2026 →Sep 20, 2026 →Sep 13, 2026 →