Discussion starter: an entity SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- AI Search
- Entity SEO
Your brand keeps getting confused with a similarly named company. What has actually disambiguated you in Google and in AI answers?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
What disambiguates a brand is rarely one tactic. It is a consistent set of attributes that the other company cannot share with you (legal name, location, category, registration identifiers, products, people), stated on your own site in a form machines can parse, and then repeated by independent sources that Google and AI systems already trust. Structured data and a claimed knowledge panel help Google attach those attributes to the right entity. For AI answers, the stronger lever is what gets retrieved when someone asks about you: if the top pages for your name describe the other company, the answer usually will too.
The main qualification is relative prominence. If the other company is much larger and owns the bare name in most searchers’ minds, no markup will make Google treat an unqualified search as being about you. In that situation the realistic goal is narrower: own “your name plus category or location”, stop your facts being merged with theirs, and make the distinction explicit in the places AI systems read.
First work out which kind of confusion you have
“Confused with another company” covers several different failures with different causes. Search your bare name, your name plus category and your name plus location in a clean browser session, ask the same questions in the AI tools your customers use, and match what you see to this table before changing anything.
Symptom Likely cause First fix Their knowledge panel appears for your bare name Google associates the name with the more prominent entity Build your own entity signals and target qualified queries; do not expect to win the bare name One panel mixes both companies (your logo, their founding date) Facts attributed to the wrong entity, or two entities merged Knowledge panel feedback with evidence links; claim the panel if you can; fix the sources repeating the error Wrong site name or logo in results Ambiguous home page signals or third-party references WebSite and Organization structured data, consistent title and og:site_nameAI answer blends facts from both Model knowledge conflates the names, or retrieved pages mention both Explicit disambiguation passages; third-party sources that separate you AI answer describes only the other company Retrieval returns their pages first; your entity is thinly represented Rank for qualified queries; earn independent coverage with citable facts Customers contact you for the other company Real demand confusion, visible in Search Console queries Clear messaging; trade mark advice if your markets overlap Why the name on its own never disambiguates anything
Picture two delegates called Sam Lee at the same conference. The name on the badge is useless for telling them apart. The organisers separate them by the company printed underneath, the session each one is presenting and the email address on the registration. If one Sam Lee has a detailed speaker profile and the other has only a name, anyone asking for “Sam Lee” gets sent to the first.
The technical version is similar. Google’s Knowledge Graph stores entities with their own identifiers and attributes, and mentions across the web are linked to one entity or another using context: co-occurring places, categories, people, links and identifiers. Large language models have no stable entity ID in their trained knowledge, so a name seen with two sets of context can blur into one description. When an AI tool grounds its answer in live search results, the retrieved pages drive much of the answer, which is why the ranking problem and the AI problem are connected.
What Google’s documentation supports
This is the strongest evidence available, because it comes directly from Google.
- Organization structured data with identifiers. Google’s Organization markup documentation says some properties are used behind the scenes to disambiguate your organisation from others, and names the ISO 6523 and NAICS properties as examples. It supports
legalName,alternateName,sameAs,addressandlogo, plus identifiers such asduns,leiCode,iso6523Code,vatID,taxIDandglobalLocationNumber. Google recommends placing it on the home page or a single page that describes the organisation. Identifiers matter most here, because they are the one attribute the other company cannot also have. - Site name signals. Google’s site names documentation says site name selection is automated and uses WebSite structured data on the home page, plus
og:site_name, the title element, headings and references to the site elsewhere on the web. It also warns that generic names are unlikely to be chosen unless they are very well-recognised brands. If Google shows a generic label or the wrong company’s name, make these signals agree with each other first. - Knowledge panel claim and feedback. Google’s knowledge panel help lets an organisation claim a panel by signing in to an official profile Google associates with it (YouTube, Search Console and certain social profiles are listed), then suggest changes. Not every panel is claimable, and local businesses are directed to Google Business Profile instead. Where you cannot claim, the Feedback link is the route, and Google asks for links to pages that verify the correction. Merged panels are fixed with evidence, so the quality of those supporting pages matters as much as the request.
- Google Business Profile. For a business with a physical location or service area, a verified address, category and phone number is often the clearest separator Google has.
Important distinction:
Markup helps Google connect attributes to the right entity. It does not create prominence. Google’s May 2026 guide to generative AI features says no special schema is needed for AI features, so treat Organization markup as an entity-clarity tool for Search, not as a lever for AI inclusion.
What helps in AI answers, and how strong the evidence is
Public evidence here is thin. No controlled public study isolates which signals stop an AI assistant from conflating two companies. The actions below follow from how these systems are documented to retrieve and ground answers, so they are worth testing rather than assuming.
Action Why it plausibly works Evidence level Publish an explicit disambiguation passage on your About page and relevant FAQs (“X is an Australian valve manufacturer founded in 2004. It is not affiliated with X Software, the US payroll company.”) Retrieval systems pull passages. A sentence that states the distinction gives the model something to ground on instead of guessing. Mechanism-based; low cost and low risk Use one descriptor everywhere: name plus category plus location Repeated co-occurrence of the same attributes is how entity linking, and model training, tie facts to a name Mechanism-based Earn third-party descriptions that separate you: trade press, industry directories, company registries, Wikidata if you meet its notability rules, Wikipedia only if you are genuinely notable AI answers draw heavily on sources other than your own site; independent pages that describe you correctly add weight to the right version Mechanism-based plus practitioner observation Rank for qualified queries (name plus product, name plus location) OpenAI documents that ChatGPT search uses its own OAI-SearchBot crawler and sometimes partners with third-party search providers; Google’s AI features draw on its index. Pages that rank for the prompt are more likely to be retrieved. Documented architecture, but the overlap between rankings and AI Overview citations varies widely by study Correct third-party pages that merge the two companies A directory or article that mixes you up is a retrievable source for the same mistake Mechanism-based Expect two different speeds. Answers grounded in live retrieval can change within weeks once the retrieved pages change. Answers drawn from a model’s trained knowledge can repeat an old confusion until that model is retrained, which you cannot schedule or influence directly.
Watch for this:
Some teams try to “win” by publishing pages about the other company or seeding forum mentions. Google’s generative AI guide says seeking inauthentic mentions is not helpful, and a page that puts both names in the same passage can increase the conflation you are trying to remove.
Hypothetical example:
“Northwind” is a 40-person Australian fire safety equipment supplier. A much larger US software company is also called Northwind. A search for the bare name shows the software company’s panel, and an AI assistant describes Northwind as a software firm with a Melbourne office, which is a blend of both.
A sensible plan accepts that the bare name belongs to the larger entity. It adds Organization markup with registration and tax identifiers, a plain disambiguation sentence on the About page, a verified Business Profile and consistent trade association and registry listings. Success is measured on “Northwind fire safety” and “Northwind Melbourne” prompts, because those are the words this company’s buyers actually use.
How to tell whether the confusion is shrinking
- Fixed prompt panel. Write 20 to 30 prompts: bare name, name plus category, name plus location, “who owns [name]”, “[name] reviews”, “is [name] the same as [other company]”. Run each several times per platform, because answers vary between runs, and record each result as correct, blended or wrong entity. Repeat monthly.
- Search Console demand signals. Use the branded queries filter and the query table to find queries that belong to the other company (their product names, “login”, support terms). High impressions with near-zero clicks on those queries is the demand-side measure of confusion.
- Retrieval check. For each qualified prompt, look at which pages rank and what they say about you. Bing Webmaster Tools’ AI Performance report (public preview) shows grounding queries and cited pages across Copilot, which tells you whether your pages are being cited for your own brand queries.
- Entity check. The Knowledge Graph Search API returns the entities and short descriptions Google holds for a name. If there is no separate entity for you, fixing panel details is premature; building third-party references comes first.
When the honest answer is a naming decision
If the other company operates in your market and category, this is not only an SEO problem. It costs enquiries and misdirects reviews, and it is a question for a trade mark lawyer as much as for search. If the markets are different, co-existence is normal, and a permanent descriptor in the brand (“Northwind Fire Safety” rather than “Northwind”) often does more than any technical fix, because it changes the words journalists, customers and directories use. That change has real costs (signage, domain, redirects), so weigh it against the measured size of the confusion, not against how irritating it feels.
If this were my brand, I would spend the first month on identifiers, the About page passage, Business Profile and the handful of third-party sources that describe the category, then judge progress on qualified prompts over the following quarter. If those come back clean while the bare name stays with the bigger company, that is a good result, and further spending on the bare name is usually money chasing a query you were never going to own.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
- Organization structured data with identifiers. Google’s Organization markup documentation says some properties are used behind the scenes to disambiguate your organisation from others, and names the ISO 6523 and NAICS properties as examples. It supports
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.