Entity resolution is the process by which a search engine or AI system decides that different mentions across the web refer to the same real company, person or product. When it works, your website, LinkedIn page, directory listings and press mentions all add up to one clear picture. When it fails, the information is split or mixed up with someone else, and AI assistants either leave you out or describe you wrongly.
Why it matters more in AI search
A traditional search result links to a page and lets the reader judge it. An AI assistant writes a sentence about your company. To do that confidently it needs to know which facts belong to you. If it finds two spellings of your name, three addresses and a similarly named company in another city, it may merge them, pick the wrong one, or avoid naming you at all.
What causes resolution to fail
- Different spellings of the company or founder names across sites
- Old addresses or phone numbers left on directories after a move
- A brand name shared with another company, product or place
- No page on your own site that states the basic facts clearly
- Profiles that exist but do not link to each other or to your website
We saw the first problem while building this site. One co-founder’s surname appeared in two spellings across different pages and profiles. For a search engine those are two different strings, and nothing tells it they are the same person until the spelling is made consistent and the profiles are linked.
The tools for fixing it
Consistent naming
Choose one exact form of the company name, founder names, address and phone number. Use it on your website, Google Business Profile, LinkedIn, directories and in press material. This is the basis of NAP (name, address, phone) consistency in local SEO, and it matters just as much for AI systems.
sameAs in your schema
The sameAs property in JSON-LD lists URLs that describe the same entity elsewhere: your LinkedIn company page, Crunchbase profile, Wikidata item, YouTube channel. Person nodes can do the same for founders and authors. This gives machines a direct statement that these profiles belong together. The JSON-LD guide shows how to structure it.
Links back from your profiles
sameAs is a claim made by your own site. Links from those profiles back to your website confirm it from the other side. Make sure every official profile links to your domain.
Wikidata
Wikidata is a free, structured knowledge base run by the Wikimedia Foundation. Each entity has an identifier beginning with Q, and its data is widely reused, including by search engines and in the datasets language models learn from. An accurate Wikidata item with official website, founders, location and identifiers is one of the strongest entity signals a company can have.
Wikidata has notability rules. An item needs to refer to an entity that can be described with serious, publicly available references, or be needed to connect other items. A new company with no independent coverage may not qualify, and items created without references are often deleted. We check eligibility before recommending it.
Google’s Knowledge Graph
Google keeps its own database of entities, visible as knowledge panels in search. You can check whether Google recognises your company as an entity by searching for its exact name and looking for a panel, or by querying the Knowledge Graph Search API. Businesses with a Google Business Profile often get a local panel first; a broader panel usually follows consistent coverage across trusted sources.
A practical sequence
- Write down the exact official form of every name, the address and the phone number.
- List every profile and listing that mentions the company or its founders, and correct any that differ.
- Add or fix an About page that states the basic facts in plain text.
- Add Organization and Person schema with sameAs links to the corrected profiles.
- Check each profile links back to the website.
- If the company meets Wikidata’s notability rules, create or update its item with references.
- Ask AI assistants what they know about the company, and trace any wrong facts to their source.
Frequently asked questions
Do I need a Wikipedia article?
No. Wikipedia has much stricter notability rules than Wikidata, and writing about your own company there is a conflict of interest under Wikipedia’s policies. Wikidata and consistent profiles do most of the work.
How long does it take for AI assistants to update?
Assistants that search the web can reflect corrected pages once their crawlers revisit them. Knowledge a model learned in training only changes when the model is retrained, which can take months.