Cited but Not Named: When AI Uses Your Page and Recommends a Competitor

Why does AI cite my page but recommend my competitor? Because citing and recommending are two separate decisions: the engine uses your page, or a review that ranks you first, as evidence for one fact in its answer, then picks the brands to name from wherever brand names are most common and most clearly tied to the question. If your name isn't sitting next to the fact it borrowed, or the cited page also names better-known brands, you get the footnote and they get the recommendation.

The pattern is well known in B2B software. Lily Ray's June 2026 study of 100 "best [category] software" queries in Google's AI Overviews found that when a brand's own self-ranked listicle was cited, the brand was left out of the recommendation 69% of the time (224 of 323 citations). Agencies have started naming it; Digital Commerce Partners calls it a "Ghost Ranking".

The online-store version is different, and in some ways worse. In the reports we run for small e-commerce brands, the cited page usually isn't a self-serving list. It's an independent magazine that tested the product and gave it an award, or the brand's own help video. The brand did the work. The answer still named other brands. In our study of 404 buyer questions for 27 small stores, this happened in 31 answers across 15 of the 27 stores. Below are four cases, checked against the raw AI answers, and what to change.

Four cases from real reports

All four come from Google AI Overview answers in reports we ran in late September 2026. Stores are anonymized; we re-read each answer and opened each cited page before writing it down.

The tent maker's own video

Asked "what size bell tent do I need for a family of four", Google's AI Overview answered "a 5-meter bell tent" and listed sources, including the brand's YouTube video comparing its 4, 5 and 6 meter tents. The video's title starts with the brand name. The answer named no brand at all. The question was about sizing, the video was used as sizing evidence, and nothing in the answer asked for a product. The brand built exactly the right content and got a footnote.

The granola a magazine picked

For "highest-rated low-sugar granola for everyday snacking", Google's AI Overview recommended KIND and Ratio, alongside product cards linked to Google Shopping listings. It also cited a national food magazine's granola roundup, whose "best paleo granola" pick is the brand, but used it to support a different grain-free granola. The brand wasn't named. The magazine had already made the recommendation; the answer kept the source and dropped the pick.

The torque wrench rated "best budget"

For "recommend bike torque tools for carbon road bikes", Google's AI Overview cited a cycling magazine's torque wrench test five times. That test names the brand's ratchet "best budget" and its preset tool "best preset". The answer recommended Park Tool, Silca, Effetto Mariposa and Topeak, all of which also appear in the same test, and most of which were backed by Google Shopping listings. The brand appears twice in the source and zero times in the answer.

The bike maker in the history article

For "are beam triathlon bikes faster than traditional tri frames", Google's AI Overview built its answer on a triathlon magazine's history of beam bikes, which credits this brand, alongside Cervélo, with bringing the design back, and then offered to compare models "like a Cervélo P5 vs. a TriRig or Quintana Roo". The source credits the brand; the answer names three others.

Why it happens

We can't see inside the models, but the four cases share a few features that explain most of it.

• The answer borrows a fact, not a verdict. Google's guide to its AI features describes grounding: the system retrieves pages and uses specific information from them to support the response. A size, a torque range or a date is easy to lift from a page. The page's opinion about which product to buy is a separate thing, and the answer doesn't have to carry it over.
• Brand picks follow broad coverage. Lily Ray found that brands recommended in AI Overviews had far more referring domains and far more mentions across AI answers than brands that were only cited. Park Tool and KIND appear on thousands of pages. One magazine award, however good, is one vote.
• The cited page names your competitors too. A roundup or group test lists eight products. If the others are better known, the answer can cite the page and pick from its list without picking you.
• Product cards come from shopping data. In the granola and torque answers, the recommended items were linked to Google Shopping listings. A brand that isn't a clean product card in Google's shopping data is harder to recommend as a product, however well a magazine reviewed it.
• Product names that don't carry the brand. A flavor name or model code on its own ("Morning Crunch", "X2") can be mentioned without the brand being attached, and then it's easy to drop.

What to change

1. Find every answer where you're a source but not a name

For each buyer question you track, record two lists: the pages cited and the brands named. Then open each cited page and look for your brand, including reviews, roundups, forum threads and videos you don't own. The cases above all came from this step; none of them were visible from "were we mentioned, yes or no". If you're starting from zero, our guide on tracking AI search citations covers the manual method.

2. Put your brand next to the fact the answer borrowed

Look at what the answer used your page for, and write the sentence that joins the fact to your product. The tent maker's sizing video becomes a size guide on the product page that says, in plain text, "Our 5 m bell tent fits a family of four: a queen air bed, two cots and room for gear." The video can carry the same sentence in its description and spoken script. The fact the engine wants and the product that delivers it now sit together, on a page the engine already reads.

This isn't writing "for AI". Google's own guidance says you don't need to rewrite content just for AI systems and should focus on what visitors find useful. A shopper reading a size guide also wants to know which tent to buy.

3. Use the reviews you already earned

The granola and torque brands had the hardest part done: an independent publication tested the product and gave it an award. Make that verdict travel with your name:

• Quote the award on the product page with the publication's name and a link to the review ("Named best paleo granola by [publication], 2026").
• Ask the publication to update the review when product names, links or prices change, so the page stays current.
• Earn more coverage the same way: send products to reviewers and communities in your niche. Google's guidance says seeking inauthentic mentions isn't as helpful as it might seem, so this means real reviews, not placed ones.

4. Name products brand-first, everywhere

Use "Brand + product + category" in page titles, H1s, feed titles and the names you give reviewers: "Brand Paleo Granola, Cinnamon", not "Morning Crunch". When a magazine, a marketplace or an AI answer copies the name, the brand comes with it. Retailer and marketplace titles often drop the brand; check yours and ask for it back.

5. Be a product card in Google's answers

Google's guide to AI features says Merchant Center feeds help products show up in AI responses. Keep a feed live with your own domain in the link attribute and complete titles, prices and identifiers, and add Product structured data to the pages it points to (see our product schema guide). For how Google picks sources in the first place, see how Google AI Overviews work.

6. Make your brand a clear entity

Add Organization structured data to your homepage with your url, logo and sameAs links to your social and marketplace profiles; Google says this helps it disambiguate your organization. Describe the brand the same way everywhere ("Brand, a small-batch maker of grain-free granola"). An engine that knows exactly who you are and what you make has an easier time attaching a borrowed fact to you.

7. Publish comparisons, not self-ranked lists

Comparison content works when it's honest. A page that compares your tent with two named alternatives on floor area, weight and price, and says who each one suits, gives the engine facts with your name attached. A "10 best bell tents" list with you at #1 is the format Lily Ray found mostly helped the competitors it listed. If a competitor wins on a point, say so; it's the part a reader and an engine are most likely to believe.

How to measure progress

Count three things per question, per engine, every month: whether you're named, whether one of your pages is cited, and whether a page that mentions you is cited. The gap between the second and third and the first is the work. When a fix lands, you'll see a cited-but-not-named question turn into a named one.

SearchMention reports list "Pages AI used that mention you" for each run: reviews, videos and roundups that mention your brand and were used as sources, including in answers that didn't name you. Most AI visibility tools list cited sources; whichever you use, check that it lets you see cited and named separately.

Frequently asked questions