How often does AI cite review pages?

A complete review or reviews label appeared in the page title or address for 221,395 of 7,434,246 cited-page links, or 3.0%. Those links covered 94,592 pages on 19,367 websites.

G2 found that 45% of surveyed B2B software buyers considered a review-site citation the strongest confidence signal in an AI answer. That survey measures buyer trust, not how often AI cites review pages. The observed citation share is smaller and applies across industries. Track review evidence as one source layer rather than treating it as the whole evidence base.

At least one review-labeled page appeared in 140,354 of 701,353 cited answers, or 20.0%. At least one appeared somewhere in the observed history for 15,052 of 18,107 prompts, or 83.1%.

Availability across repeated answers is not the same as inclusion in the next answer. Report answer reach and prompt reach separately.

ChatGPT Search used 53,435 review-labeled links among 1,663,858 cited-page links, or 3.2%. Google AI Mode used 167,960 of 5,770,388, or 2.9%. The rate ratio was 1.1 to 1, and the gap was 0.30 percentage points.

Keep an engine field in the audit, but do not overstate this gap. It is much smaller than the differences between prompt types, website categories, and industries in the same cut.

Review-labeled pages supplied 16,157 of 108,359 links on prompts that named reviews, ratings, testimonials, or feedback, or 14.9%. They supplied 205,238 of 7,325,887 links on other prompts, or 2.8%. The first rate was 5.3 times the second.

Compare review-page rates within prompt intent. A review strategy should start with the buyer questions that explicitly ask for peer judgment, not an aggregate target applied to every prompt.

Review-labeled pages supplied 36,888 of 390,186 editorial links, or 9.5%. News recorded 6.9%, general websites 4.3%, directories 2.1%, SaaS websites 1.1%, and documentation websites 0.55%.

Review evidence is not limited to directory profiles. Compare page labels within a website category before deciding which publishers to inspect. These broad categories do not judge page quality or factual support.

YouTube supplied 13,462 review-labeled links, TechRadar 10,980, Gartner 10,159, The New York Times 10,158, RTINGS 5,827, and G2 3,651. The top ten websites together supplied 63,556 of 221,395 review-labeled links, or 28.7%.

Seer found Trustpilot was the most-cited review site inside a 1,926-brand review-profile study. This broader corpus asks a different question and includes editorial reviews, videos, product testing, and review directories. Build a publisher map from the prompts and pages in your market instead of assuming one directory is the universal gatekeeper.

Consumer Electronics recorded 11,915 review-labeled links among 107,769 links, or 11.1%. Sports recorded 7.1%, Food and Beverage 6.0%, and Consumer Goods 5.6%. Data and Analytics recorded 1.7%, and Events 1.4%. The Consumer Electronics rate was 7.8 times the Events rate.

Use an industry baseline before setting a review-page target. These differences describe the observed prompt and source mix and do not show that an industry caused a citation.

TechRadar's Ahrefs review supplied 1,163 cited-page links. Its Dynatrace review supplied 975. Position Digital's AI visibility tools review supplied 442, The Good Trade's sustainable jeans review supplied 403, and ICAgile's project-management tools review supplied 402. Three more cleaned page titles complete the leaderboard.

Review the exact page and buyer question. The leaders include single-product tests, category shortlists, and consumer guides. A review-page total does not show which format is useful for one market.

The combined title-or-address rule returned 221,395 of 7,434,246 links, or 3.0%. Title labels returned 177,063, or 2.4%; address labels returned 140,546, or 1.9%; and 96,214 links carried both. Excluding academic websites and common research or business-review phrases left 218,040 links, or 2.9%.

The rule misses review content without a review label and includes scholarly reviews, annual reviews, and pages where review is used in another sense. It does not measure ratings, page quality, factual support, traffic, causation, or whether the page changed an AI recommendation.

Review-labeled pages supplied 221,395 of 7,434,246 cited-page links. The rate rose to 14.9% on review-focused prompts, and the top ten websites supplied only 28.7% of review-labeled links.

Keep the cited URL, engine, prompt, page label, and publisher category together. Audit review-focused prompts first. Separate product tests, category shortlists, editorial reviews, videos, and directory profiles. Do not infer citation lift from creating a profile or adding review to a title. Rerun the fixed method next quarter before calling any difference movement.