AI can describe your business and still never recommend it.

Generative AI platforms can recognize a brand and still fail to recommend it, according to a controlled study that Search Engine Land reported this week.

What the study found

Search Engine Land published an article by Maryanna Franco on October 15, 2026 describing a study of 14,140 controlled API runs across 12 brands and five generative AI platforms. INDEPENDENT RESEARCH It belongs to the wider shift we cover under AI search optimization. The core finding is that a model can know a brand and still fail to recommend it. The sharpest example is TALA, a U.K. activewear brand: the language models placed it correctly in athleisure when the category was introduced explicitly, while Google’s Knowledge Graph described TALA as a financial services company. It appeared in none of the tested athleisure or sportswear recommendation responses. INDEPENDENT RESEARCH

The second example cuts the other way. New Balance had the highest Knowledge Graph result score in the sample and a recognition score of 1.0, yet received zero athleisure recommendations, because its Knowledge Graph description identified it as a footwear company and the models categorized it as athletic footwear. INDEPENDENT RESEARCH So the problem is not only a missing record. A strong record that points at the wrong category fails too.

The fixes the study recommends are plain: disambiguate organizations that share a name, make each one’s identity, attributes and relationships explicit and consistent, strengthen the entity home, reconcile organization details and aliases across authoritative profiles, get corroboration from relevant third parties, and submit corrections where Google’s Knowledge Graph holds a wrong or conflated record. INDEPENDENT RESEARCH

Two limits before you take this to a local business. The sample is apparel and activewear brands, not plumbers, and it is one study. Applying it to local service businesses is our inference, not a finding. DHM INTERPRETATION Google’s own guidance also says its AI features are rooted in core Search ranking and quality systems and need no special markup, so none of this is a shortcut. GOOGLE DOCUMENTED

What an entity is for a local service business

An entity is the record systems keep about who you are. For a local business it is assembled from your Google Business Profile (name, primary category, address or service area, phone), your website (the About page, the schema on it, the people named), the profiles you link as the same business, directory listings, and what reviews and other sites say about you. When those agree, a model has one clear thing to describe and recommend. When they disagree, it hedges or picks someone else. The profile half of that record is the Google Business Profile, and keeping the facts identical everywhere is the subject of our guide to business info consistency.

What we found when we checked ourselves

We ran the check on Digital Handshake Media on October 16, 2026, using the Knowledge Graph and search tools in our SEO Master connector, Search Console and DataForSEO. DHM DATA

Name collision. Searching our exact name from Canada puts our website first, with a knowledge panel typed Internet marketing service. Two of the top ten results are ours. The other eight are a book (listed twice), a podcast, two sites with near-identical names such as Handshake Digital, and two blog posts and a social post that use “digital handshake” as a metaphor. DHM DATA That is a collision a model has to resolve when someone asks.

Person versus company. Google’s public Knowledge Graph Search API returned no entity for “Digital Handshake Media” and one Person entity for Omar Alocozy, with no description attached. DHM DATA Treat that carefully: the API is a partial public view, not Google’s internal graph, and the Business Profile does produce a knowledge panel. It does show the shape of the gap the study describes, a founder the graph recognizes and a company name it does not return.

Name variants in real searches. In Search Console from July 16 to October 13, 2026, the homepage query with the most clicks was “digital handshake mmc” (5 clicks from 23 impressions, average position 5.8). Other variants appear too, such as “handshake digital” and “handshake marketing”. DHM DATA Searchers use several versions of the name, and a model has to map all of them to one company.

Fact drift, caught. Earlier this month an audit run through ChatGPT and Gemini flagged contradictory review counts and category data across our own properties. DHM DATA Our response was to keep every entity fact in one source file and make the build fail when any page disagrees with it. You do not need a build step to copy the idea: keep one written list of your facts and check every listing against it.

Search interest in the topic is also measurable. Google Ads data via DataForSEO shows “google knowledge panel” averaging about 1,900 US searches a month over the last year and 2,400 a month in every month from July to September 2026, up from 1,600 in the winter months. DHM DATA

The five-minute entity check

1. Search your exact business name in quotes in a private window, once with your city and once without. Write down who else appears in the top ten.
2. Read your own knowledge panel or listing. Check the category it shows against the service you want to be recommended for.
3. Ask two questions in an assistant: “What is [your business]?” and “Who is the best [your service] in [your city]?” The study’s findings suggest a mismatch when the first answer is right but the second leaves you out.
4. Repeat each question three to five times. AI answers vary run to run, so track how often you appear, not whether you did once.
5. Compare your facts. Name, primary category, address or service area, phone and hours should match on your profile, website and top listings.

Fixing it

• Pick one name and one primary category and make your Business Profile, website, schema and listings match exactly.
• Build an entity home. Your About page should say who you are, where you operate and who runs the business, and link your official profiles. Add Organization and Person markup that names both.
• Add a plain disambiguation line if another business shares your name, for example “not affiliated with…”. Keep it factual.
• Earn corroboration. A chamber of commerce, industry directories and your founder’s professional profile can reinforce the same facts from outside your own site.
• Request corrections where a record is wrong, through your Business Profile edits or the feedback on a knowledge panel. You cannot force a change. Keep your own facts consistent while you wait.
• Re-test monthly with the same prompts and the same number of runs, and keep the log.

This work takes time, and no one can guarantee an assistant will recommend you. It is one part of AI visibility you can control. Once an assistant recommends you, the next question is whether you can measure the resulting visits. Measuring AI referrals without UTM tags covers that, and this week’s Update explains why it matters now. DHM INTERPRETATION