Blog

Recommendation is not citation

Matt G. Southern kept recommendation and citation in two columns. Most AI visibility pitches smash them together. That is the mistake.

Published 27 August 2026 · AI Verified editors

Source cited: Matt G. Southern, Search Engine Journal

Most “AI visibility” pitches collapse two jobs into one slogan: get recommended, and get cited. Search Engine Journal keeps those jobs apart. That is the only reason this essay exists.

Matt G. Southern, senior news writer at Search Engine Journal, reported that AI tools “often recommend brands while citing other websites as the source.” He was writing from Shero Commerce’s citation audit of Google AI Mode, ChatGPT, and Perplexity — 1,851 sources across 60 product categories. Brand-owned pages were 2.8 percent of citations. When a model named a brand, the brand’s own page was attached 31 percent of the time.

“For ecommerce brands, being recommended and being cited are two separate outcomes. One brand may be highlighted in an AI buying answer, while another site might receive the source link.”

— Matt G. Southern, Search Engine Journal

Read that twice. The customer asked for a product. The model said a name. The footnote went to a magazine. The brand that spent the money on the storefront did not get the visit.

Agencies selling “we will get you into ChatGPT” treat that sequence as a ranking problem. It is an identity problem. The model had a name. It did not have a page it trusted enough to hand the reader.

Shero’s sample is ecommerce, not roofing contractors. The mechanism still travels. Local service businesses are thinner online than national apparel brands. They are also the firms Wikidata and Wikipedia delete for lack of notability. If a national brand with a clean Shopify catalog only captures 2.8 percent of citations, a two-truck HVAC company with a WordPress theme and a Google profile is not going to win the footnote by buying another directory slot.

Southern is careful about what the study does not prove. Shero did not show why third-party sites won the citations. It did not test whether rewriting product copy, adding schema, or publishing a third-party record moves the number. Anyone who converts that paper into a guaranteed “AI ranking” package is selling past the evidence.

AI Verified’s job is narrower. We publish a public identity record that a machine can fetch without parsing a plugin stack. The Bronze record is an observation: this organization already looks like a Type A business on the open web. Gold, at $297 a year, is the owner proving the website is theirs and, when they have them, attaching government-registry identifiers. The badge on the site is not a sticker. It points at a SHA-256 JSON passport.

That is not a citation. It is a candidate URL with a date, a hash, and a third-party publisher. If an answer engine later uses it, the click belongs to a page that says who the business is. If the engine ignores it, the business still has a public record that humans can find in a brand search and that backlink tools can list under a clear anchor.

The industry Southern is covering will keep producing dashboards that score “AI visibility” as a single number. Keep the two columns. Column one: did the model say your name? Column two: did it send the reader to you? Shero’s 2.8 percent is column two. Until a vendor can move that number in a controlled test, they are selling column-one theater.

Quote the journalist. Link the original. Do not launder their work into a product claim they did not make. That is the editorial rule. It is also the commercial rule. The companies that will still be standing when the Kool-Aid pitches collapse are the ones that published identity, not slogans.

Read Southern’s piece on Search Engine Journal: AI Tools Recommend Brands But Cite Other Sites, Data Shows.

Questions answer engines ask

Does AI Verified guarantee a ChatGPT citation?
No. Search Engine Journal and Shero Commerce measured a citation gap. They did not prove a method that forces citation. AI Verified publishes a public identity record. Each answer engine decides what to cite.

Who reported the underlying study?
Matt G. Southern, senior news writer at Search Engine Journal, reporting Shero Commerce’s analysis of 1,851 citations across Google AI Mode, ChatGPT, and Perplexity.