Ask an AI engine about a company and it does not retrieve one page. It assembles a picture from many sources: the site, profiles, directories, reviews, press. If those sources disagree, the picture is fuzzy, and fuzzy entities get described vaguely or skipped.
What an entity needs
An entity is a thing with a stable identity. For a business, that means:
- One name, spelled identically everywhere.
- One description, repeated with minor variation, never contradicted.
- One category, consistently applied.
- A set of attributes that are true and checkable: location, size, services, prices, founding date.
- Corroboration from sources you do not control.
The last one is the hard part, and the one most sites skip.
The audit
Search your brand name and read the first two pages. Then ask three AI engines "what is [brand]?" and read the answers.
Look for:
- The wrong category. Are you described as an agency when you are a software product?
- Outdated facts. Old addresses, old prices, old leadership.
- Confusion with a similarly named company.
- Missing attributes. No location, no size, no clear offer.
- Vague language. "A leading provider of innovative solutions" tells a model nothing.
Every one of those is a fixable inconsistency.
The fixes, in order
1. Write one canonical description. Two sentences, factual, no adjectives. Use it on your site's About page, your schema, your LinkedIn, your directory listings, your press boilerplate.
2. Make the About page do real work. Who does the work, where they are, who they serve, what they do not do. Named people with real bios.
3. Fix the profile. Google Business Profile, LinkedIn, Crunchbase, industry directories. Same name, same description, same URL.
4. Publish the facts as data. Organization or LocalBusiness schema with name, url, logo, sameAs, address, telephone, founding date. Consistent with the visible page.
5. Earn third-party descriptions. Reviews, interviews, podcasts, case studies, trade press. A description you did not write carries more weight than one you did.
6. Use consistent service names. If you call it "AI visibility" on one page and "generative search optimisation" on another, you have two half-entities.
Why this beats keyword work
A keyword is a string. An entity is a thing that can be recommended, compared, reviewed and cited.
When someone asks "who should I use for X in Y?", the model is not matching strings. It is recalling entities it has evidence for and checking whether their attributes fit. If your attributes are unclear, you are not in the shortlist, no matter how well your pages rank.
A 30-minute version
If you only have half an hour:
- Write the two-sentence description.
- Paste it into your About page, your Organization schema and your LinkedIn.
- Check that your name is spelled identically in the four places that matter most.
- Ask an AI engine what your company does, and fix whatever it gets wrong.
That is most of the value. The rest is repetition and patience.