How to correct what AI search says about you
Ask an AI assistant about yourself or your company and you'll often get a confident answer that's mostly right and partly invented. You can't edit the answer directly. You can change what the assistant reads, and that changes what it says.
Why the errors happen
AI answers are stitched together from whatever the system retrieves plus whatever it absorbed in training. Three failure patterns cause most of the mistakes I see:
- Name collisions. Two people or companies share a name, and the assistant blends them. A search for one marketing consultant can pull in a theatre director and an ad-network salesperson with the same name.
- Gap filling. When sources are thin, the model fills holes with plausible details. A neighborhood called Salt Lake in Honolulu becomes Salt Lake City, Utah, because the city is the more common phrase.
- Stale sources. An old bio, an expired company page, or a ten-year-old article outranks the current truth because nothing newer says otherwise.
Step 1: Collect the errors properly
Run the same handful of questions across ChatGPT, Perplexity, Gemini, Claude, Copilot, and Google with AI Overviews and AI Mode. Ask the plain questions a stranger would ask: who is this, what do they do, where are they based, what have they done. Save each answer with the date and the sources it cites. Run the set two or three times, because answers vary between runs, and an error that shows up every time matters more than a one-off.
Step 2: Trace each error to a source
Most assistants now show citations. Follow them. For each wrong fact, sort it into one of three buckets:
- A real source says it. Something on the web states the wrong fact. That source has to be fixed or outweighed.
- A real source almost says it. The source is right, but ambiguous, and the assistant misread it. "Salt Lake" without "Honolulu" next to it is an invitation to guess.
- Nothing says it. The assistant invented the detail to fill a gap. The fix is to fill the gap yourself.
Step 3: Fix what you control first
Your own pages carry the most weight, because they're the most authoritative source about you. On your About page:
- State the correct facts in plain sentences, with the specific words that remove ambiguity: the full place name, the full company name, the exact role.
- Where a mistake is common, correct it explicitly. A line like "Salt Lake, the Honolulu neighborhood, not the city in Utah" gives every crawler the correction in your own words.
- Add Person or Organization structured data with your name, alternate names, job title, location, and links to your official profiles.
- Keep the page fast and in plain HTML, so crawlers that don't run JavaScript still read every word.
Step 4: Fix or outweigh other sources
For wrong facts on sites you don't control:
- Profiles you own: update LinkedIn, Upwork, Crunchbase, author bios, speaker pages, and directory listings so they agree with each other.
- Publications: ask editors to correct factual errors. Most will fix a wrong title or location when you ask politely with evidence.
- Sources you can't change: publish newer, clearer material that states the correct fact, and earn links to it. Assistants weigh recent, consistent, well-linked sources over a single old page.
Step 5: Separate yourself from your namesakes
If you share a name, make your distinguishing details unmissable: your field, your city, your middle name or initial, and the profiles that are yours. In structured data, use sameAs links to your own profiles and nothing else. Consistent pairing of your name with your field ("Christopher Abraham, SEO consultant") teaches the systems which person you are.
Step 6: Use the feedback buttons, then wait
Every major assistant has a thumbs-down or feedback option. Use it on wrong answers about you, with a short correction and a link to your source. It won't fix anything on its own, but it adds a signal. Then give it time. Retrieval-based answers can change within days of a crawl; facts baked into a model's training change only with the next model.
Step 7: Re-run the questions monthly
Keep the same question set and run it once a month. Track which errors disappear, which persist, and which new ones appear. Persistent errors usually point to a source you haven't found yet.
What not to do
- Don't publish exaggerations to counter errors. The next assistant will repeat those too.
- Don't let unverified AI summaries back into your own bio. Check every detail against your own records before you publish it.
- Don't flood the web with thin pages repeating your name. A few strong, consistent sources beat dozens of weak ones.
For businesses, the full approach is under AI SEO, GEO, and AEO. For people, the companion guides are one person, two names and writing an About page AI trusts.