Key Takeaways
- Two separate monitoring sessions: visibility metrics (are you mentioned in discussions about your market?) and brand monitoring (what do AI systems say when asked about you?).
- What we're looking for: factual errors (offer, price, location, executive), outdated information, confusion with someone of the same name, negative comments, and their sources.
- The correction: AI systems repeat their sources; correct the source (website, LinkedIn, Google listing, directory, review, article), not the answer.
- Tempo: Monthly for brand-related issues, quarterly for a full audit, and immediately in the event of a crisis.
Assessing whether AI systems mention you in relation to issues in your market is one thing (see track its visibility on ChatGPT and Perplexity). Knowing what they say about you when asked directly is another: an assistant might describe an offer you no longer provide, place you in the wrong city, confuse you with someone who shares your name, or treat a negative review as a general truth. This brand monitoring is the AI equivalent of theonline reputation. This article outlines the questions to ask, what to look for, how to trace the issue to its source, and how to fix it.
Issues in Monitoring
| Family | Sample Questions | What We Check |
|---|---|---|
| Identity | "Who is [brand]?", "What does [brand] do?", "Where is [brand] based?" | Description, category, location, years in operation, manager |
| Offer and Price | "How much does [brand] cost?", "What does [brand] have to offer?" | Accuracy, freshness |
| Reputation | "Is [brand] reliable?", "What are people saying about [brand]?", "What are the downsides of [brand]?" | Sentiment, Sources of Criticism |
| Comparison | “[Brand] or [competitor]?”, “alternatives to [Brand]” | Perceived positioning, associated competitors |
| Recommendation | "Would you recommend [brand] for [purpose]?" | Recommendation Criteria |
| People | "Who is [the leader]?" | Accuracy, Homonyms |
About 15 questions are enough—asked each month to ChatGPT (search enabled), Perplexity, Gemini, and Google (AI mode), in incognito mode, twice each.
What to Look For
- Factual errors: wrong city, outdated data, offer no longer available, outdated price, incorrect executive. These errors often occur when the website has changed and third-party sources haven't been updated accordingly.
- Common Misconceptions: with a company of the same name, a similar brand, or a former company name.
- Information dated: The AI cites a 2022 article as current.
- Negative comments: an isolated review, an old dispute, or a forum comment taken as a general characteristic.
- Absences: AI knows nothing about you, or describes you in general terms: your identity has not been established.
- Hallucinations: fictitious details (reward, customer, feature); rare when web search is enabled, but possible.
Go back to the source
The AI assistants repeat what they read. For each error, ask for the sources (Perplexity displays them, ChatGPT lists them, Gemini provides them upon request) and identify the original page: an outdated directory, an old website page that’s still indexed, a news article, an obsolete LinkedIn profile, a review, or a forum post. When no source is cited, search for the exact wording in Google: it often comes from a directory or an aggregator. The different types of sources are described in Sources of Generative AI.
Correct
| Origin | Correction | Effective Date |
|---|---|---|
| Your website (outdated page, previous offer) | Update or redirect; improve the "About Us" page; see Optimize your "About Us" page | 2 to 6 weeks after reindexing |
| LinkedIn, Google listing, directories | Update each profile using the reference entity record | A few weeks |
| Dated news article or blog post | Request an update or a date notation; publish more recent information that will take precedence | Variable |
| An isolated negative review | Respond publicly and factually; gather recent opinions to restore balance; see reply to reviews | 1 to 3 months |
| Forum or community | Reply to the discussion if it's open, using facts; do not delete or make things up | Variable |
| Homonym | Strengthen distinctive identifiers (sector, city, alternateName, sameAs in the markup); create the Wikidata entry | 1 to 3 months |
| Hallucination with no source | Publish the exact facts in a prominent manner; notify the template editor if the damage is real | Uncertain |
There is no direct “right of reply” in search engines; the only lasting solution is to make the correct source more visible and more consistent than the incorrect one.
Pace and Organization
- Monthly: The fifteen brand-related questions, recorded in the same spreadsheet as the visibility measurement, along with the noted discrepancies.
- Quarterly: Comprehensive audit of third-party sources (directories, profiles, reviews) against the reference entity record.
- Every time there is a change: New offer, new price, relocation, change in leadership: Update all sources on the same day, then check the AI responses one month later.
- In the event of a crisis: Widespread criticism, negative articles, public disputes: weekly summary of reputation-related issues and factual responses based on sources.
Tools
GEO monitoring tools generally cover brand monitoring: they track brand-related topics, measure sentiment in mentions, and alert users to changes. Traditional online reputation monitoring tools (mentions on the web and social media) remain useful for detecting the source before it reaches AI systems. A management platform that combines both—along with traffic and rankings—prevents fragmentation. The metrics to track are in GEO KPIs.
How GreenRed Helps You
Instead of juggling multiple tools, GreenRed's GEO tracking brings these metrics together in a single dashboard, compares them over time, and shows you which actions to prioritize. You can try it for free—no credit card required—from the page Rates.
Frequently Asked Questions
Can I ask ChatGPT to correct information about my company?
Not directly: there is no editorial correction procedure for businesses, except for personal data (GDPR rights of access and rectification) and proven damages that must be reported to the publisher. The sustainable approach is to correct the source that the AI reads (website, LinkedIn, Google listing, directories, press) and make it more visible.
How often should I check what AI is saying about my brand?
Once a month for about 15 brand-related questions (brand identity, product/service offerings, pricing, reputation, comparisons); a comprehensive audit of third-party sources every quarter; and a check one month after any significant change (product/service offerings, pricing, address, executive leadership). In the event of a reputation crisis, switch to weekly monitoring.
What should I do if an AI confuses my company with another company that has the same name?
Strengthen distinctive identifiers: industry, city, and a detailed description on the “About Us” page; use the Organization schema with alternateName and sameAs links to your profiles; create a Wikidata entry; and ensure strict consistency of the name across all platforms. Confusion typically resolves within one to three months as the entity becomes more clearly distinguished from its namesake.
Can an AI make up facts about my company?
Yes, this is a misperception—one that has become less common since AI assistants began relying on web searches—but it’s still possible for companies that lack sufficient documentation. The solution is to document your organization clearly and consistently using sources that AI systems can read, so they no longer have to fill in the gaps.