Research · November 22, 2026

Which sites AI shopping answers rely on

We asked AI search 100 shopping questions across 7 categories and recorded every site it found and every site it cited. A few review sites carry most of the answers, and which ones depends on the engine and the category.

100

shopping questions, 49 general and 51 specific, across 7 categories.

483 → 181

sites the search returned, versus sites the answers actually cited.

3

review sites (Tom's Guide, RTINGS, TechRadar) were cited in 46 of 100 answers.

11 of 28

Ask Brave answers carried sponsored product tiles, 19 of 26 of them from Amazon.

Found is not the same as cited

A search returns about 12 sites per question, and the answer cites about 5 of them. Syndicated copies on Yahoo and AOL showed up constantly but were almost never the source an answer named. The sites that write and test the reviews themselves won the citations.

Found by the searchCited in the answer
Yahoo463Tom's Guide3733AOL321TechRadar2925RTINGS2827Popular Science227Digital Trends227BestReviews202CNN1412time.com146

Number of the 100 questions where each site was returned, and where it was cited. The 10 sites returned most often.

Every category has its own winners

Specialists own their category. Edmunds and Kelley Blue Book carry car questions, OutdoorGearLab carries outdoor gear and Sleep Foundation carries mattresses. Travel is the most spread out: no site was cited in more than 5 of 14 travel answers.

CategoryQuestionsMost cited2nd3rd
Electronics16Tom's Guide (14)RTINGS (12)TechRadar (8)
Home15RTINGS (12)Tom's Guide (10)TechRadar (8)
Travel14timeout.com (5)thepointsguy.com (4)fodors.com (3)
Autos14Edmunds (12)Kelley Blue Book (10)Cars.com (8)
Health and sleep14Tom's Guide (6)NBC News (5)Sleep Foundation (5)
Outdoor and apparel14OutdoorGearLab (10)REI (7)GearJunkie (6)
Software13TechRadar (8)techrepublic.com (5)zapier.com (3)

Number in brackets: answers in that category that cited the site.

General and specific questions get similar sources

A "best robot vacuum" question and a "Roborock vs Dreame" question lean on the same review sites. The top sites held up on specific questions, with brand names, budgets and comparisons, nearly as well as on general ones.

General questions (49)Specific questions (51)
Tom's Guide1815RTINGS1314TechRadar1411OutdoorGearLab94CNN93Edmunds66Sleep Foundation64

Answers citing each site, split by question type. The 7 most-cited sites overall.

A second engine sees a different web

Each AI engine searches its own index. Claude's search appears to draw on Brave's index; ChatGPT uses Bing and partner content; Google's AI Mode, AI Overviews and Gemini use Google's index; Perplexity and Brave run their own. On the 28 questions we also ran through Ask Brave, Reddit was the most-used source (9 of 28) though it never appeared in Claude's search. Forbes is the clearest swing: missing from Claude's results, listed in 7 of 28 Ask Brave answers, and ranked #1 on Gemini and Google AI Overviews in Foglift's published study.

reddit.com9CNN8New York Times8Tom's Guide7Forbes7U.S. News7CNET6Yahoo6RTINGS6PCMag6

Ask Brave: answers, out of 28, that listed each site among their sources.

Why the top sites win

  • They let AI search bots in. None of the 10 review sites we checked blocks the search bots of OpenAI, Anthropic, Google or Bing. Several, including Forbes, Tripadvisor and Edmunds, block bots that collect training data while staying open to search.
  • They show a recent update date. 6 of the 10 top pages we read carried an "updated" date from the last two months.
  • They publish their own test results. Battery hours, lab scores and measured ratings give an answer a specific fact to cite and a reason to name the source.
  • They cover many products on one page. A "best X" list likely matches both the general question and each product the agent then looks up.

The ads are already moving into the answer

ChatGPT began testing ads in answers on February 9, 2026, Google took ads in AI Mode worldwide in October 2026, and Copilot shows them too. In our Ask Brave run, 11 of 28 answers carried Sponsored product tiles, mostly from Amazon and sometimes for an older model than the one the answer recommended. None of the comparison questions carried them.

Those ads sit beside the answer, and each engine decides when to show them. The pages agents read to write the answer are a separate surface: a placement inside a cited review reaches every engine that cites it. That surface is what Agent Ready Ads sells.

What it means

  • For publishers: agent visits to your reviews are growing, and the banner and sidebar ads on those pages are invisible to them. Placements inside the text can earn from those visits.
  • For brands: agents learn about your products from review sites, not your own site. In our run, brand sites were almost never cited, even for questions that named the brand.

Method and limits

On November 22, 2026 we wrote 100 shopping questions across electronics, home, travel, autos, health and sleep, outdoor and apparel, and software, about half general and half naming a brand, budget or comparison. Each ran once through Claude's built-in web search with US results. We recorded every site returned and the 3 to 6 sources chosen to cite in the answer; citations were chosen from search results, not full pages. 28 of the questions, 4 per category, also ran through Ask Brave, where we recorded the sources listed and cited (1 to 10 per answer) and any sponsored tiles. ChatGPT, Perplexity, Gemini, Copilot and Google's AI features weren't measured directly; where we mention them we name the published source. One run on one day is a snapshot: published trackers show citation mixes shifting week to week.

Sources: Foglift, top 100 AI-cited domains, Q2 2026 · OpenAI, testing ads in ChatGPT · Google, ads in AI Mode · TechCrunch on Claude and Brave · Google, AI features and your website · Brave, Ask Brave

Want this study run on your own site, category or product list?

hello@agentreadyads.com