How often does AI cite a competitor's website?

A reviewed link connects one cited page to one ranked recommended brand that the page mentioned. The cited domain belonged to a different brand in the same market in 6,965 of 136,478 links, or 5.1034%.

Lily Ray's 69% result covered 224 of 323 self-promotional listicles that Google AI Overviews cited while leaving the publisher out of the recommendation. This study measures something different across two engines: the share of all eligible reviewed links whose cited domain belongs to another brand in the same market. It refines that result rather than contradicting it. Marketers should inspect who owns the source behind each recommendation instead of treating every cited brand site as support for its own products.

Another known brand owned 55,607 of the 136,478 reviewed links, or 40.7443%. Competitors in the same market supplied 6,965 of those 55,607 links, or 12.5254%.

The broad first-party versus third-party split misses an important distinction. A brand-owned page can be first-party for the publisher and still function as third-party evidence for the brand being recommended.

Google AI Mode used a competitor-owned page in 5,169 of 83,592 reviewed links, or 6.1836%. ChatGPT Search did so in 1,796 of 52,886, or 3.3960%.

A competitor-source audit should cover both engines. The combined rate hides a clear engine difference, but this comparison does not isolate why their source mixes differ.

Competitor websites supplied 1,840 of 37,250 reviewed links for #1 recommendations, or 4.9396%. The rate was 5.5072% for ranks #2 to #3 and 4.8946% for rank #4 or lower.

The pattern is not confined to brands near the bottom of an answer. Check who owns the sources behind the top recommendation as well as the brands further down the list.

Among markets with at least 20 competitor-owned links, Technical Recruiting Assessment Software had 37 of 211 reviewed links from competitors, or 17.5355%. Online Community Forum Platforms followed at 49 of 281, or 17.4377%. Reddit and Community Marketing Services recorded 1,475 of 8,708, or 16.9384%.

Market context changes the expected rate. Teams should Compare their source mix with their own buyer market and keep the link counts in view, because small markets can post high rates from few links.

Comparison and best-of pages supplied 5,907 of 6,965 competitor-owned links, or 84.8098%. Other pages supplied 961, customer and case-study pages supplied 67, and product and documentation pages supplied 30.

The practical risk is concentrated in pages that name several options. Audit comparison content for which competitors it helps an AI engine recommend, not only whether the page earns citations.

New Relic pages supplied 155 reviewed links for Dynatrace across 14 pages and 33 prompts. They supplied another 144 for Datadog across 17 pages and 36 prompts. FlyPix pages supplied 138 links for Planet across 20 pages and 30 prompts.

These pairs identify where to inspect the actual pages and answers. They do not show that the source owner intended to endorse the recommended brand or that the recommendation was factually correct.

The study counted each link once and found zero duplicates. It excluded 23,289 links without one clear brand owner for the cited domain and 15,617 without a primary market on both brands. Alias checks removed 136 apparent competitor links, changing the rate from 5.2030% to 5.1034%.

These rules reduce false competitor matches. The remaining sorting still depends on known domain ownership and known markets for each brand, so the result should be read as a conservative observed cut rather than a complete count of every competitor relationship.

Competitor websites supplied 6,965 of 136,478 reviewed links, the rate differed by engine, and comparison pages supplied 5,907 of the competitor-owned links. The named pairs show which source owners repeatedly support other brands in the same market.

Track the owner of every cited domain next to the brand being recommended. Split the audit by engine and recommendation rank. Review your comparison pages for competitors they support, and inspect competitor pages that support your brand. Repeat the same query over a window of the same length next quarter before treating any pair as lasting.