Entity Resolution: Teaching AI That Two Names Are One Thing
Your business probably has more than one name. The legal name on the invoice, the brand on the sign, the shortened version guests use, the misspelling that shows up in reviews, the old name from before the rebrand. To you they are obviously the same place. To a machine they are five different strings of letters, and unless something tells it otherwise, five different places. Entity resolution is the job of working out which mentions refer to the same real thing.
The problem in one example
A hotel is listed on its own site as "The Azure Hotel and Spa". Google Maps has "Azure Hotel". A travel blog writes "the Azure in Pattaya". A review says "Azur Hotel". A booking site uses "AZURE HTL PATTAYA". An AI gathering information about hotels in Pattaya sees five mentions. If it treats them as five hotels, each one has a fifth of the reviews, a fifth of the facts, and a fifth of the credibility. If it merges two of them with a different hotel called Azure in another city, it starts mixing up pools and prices that belong to someone else.
Neither outcome is good, and both happen constantly.
How resolution works
Entity resolution compares mentions and decides whether they point to the same thing. It rarely relies on the name alone. It looks for corroborating clues: the same address, the same phone number, the same website, the same city, the same opening date, the same parent company. Two mentions that share an address and a phone number are almost certainly one place, even if the names differ. Two mentions with identical names in different countries are almost certainly not.
Modern systems combine three approaches. Rules handle the obvious cases: identical website domain means identical business. Fuzzy matching handles near misses: "Azur" is one letter away from "Azure". Embeddings handle meaning: a model can judge that "the Azure in Pattaya" and "Azure Hotel and Spa, Pattaya" are describing the same thing even though the words are arranged differently.
Once the system is confident, it merges the mentions into a single entity with a single identity, and everything that was attached to any of the mentions now attaches to the one thing.
Why it matters more now
Search engines have done this for years to build their knowledge panels. Now AI assistants do it on the fly whenever someone asks a question. When a user asks "is the Azure in Pattaya good for families", the assistant has to resolve "the Azure" to a specific hotel before it can say anything useful. If the resolution fails, it either shrugs, answers about the wrong hotel, or blends several into a confident mess.
For a business, this is quietly decisive. Your reputation, your facts, and your reviews only count for you if a machine can tell they are about you.
What makes resolution easy
You cannot run the resolution for someone else's AI, but you can make it easy to get right.
Use one canonical name everywhere you control, and use it consistently. Variations are fine in prose, but the official name should appear on every page, profile and listing.
Publish the corroborating clues together: name, address, phone, website, in the same place and in the same form, on your site and on every listing. These are the matching keys machines rely on.
Give your business a stable identifier where possible. Structured data on your website can declare "this page is about this specific business" and link to the same entity on Wikidata, Google, or your own canonical page. A shared identifier ends the guessing.
Keep sub entities distinct. A hotel's restaurant, spa and bar are their own entities with their own names. Make clear they belong to the hotel, but do not let their names blur into the hotel's.
The takeaway
Entity resolution is how machines decide that two names are one thing, or that one name is two things. Every fact about you depends on it landing on the right side. Clear, consistent, corroborated identity is what makes it land.