What Does ChatGPT Say About Your Company? Audit and Fix It

October 13, 202611 min read

The only way to know what ChatGPT says about your company is to ask it the way a customer would, in a fresh session, across several assistants, and write down the answers. Brands doing this for the first time commonly find a wrong fact, a missing product line or a competitor being recommended in their place. The fix is not arguing with the chatbot. It is correcting the sources the models read, then re-testing on a schedule.

Key takeaways

Run the same 12 prompts across ChatGPT, Gemini, Perplexity, Claude and Google AI Mode, in logged-out or fresh sessions, and score every answer on four criteria.
Assistants describe you from third-party sources far more than from your website; Muck Rack’s May 2026 study found earned media made up 84% of AI citations.
Wrong answers trace back to wrong or missing sources: an outdated Wikipedia line, a stale Crunchbase entry, a review site, or a press article that no longer reflects reality.
Platform feedback tools exist and are worth using, but source correction is what changes answers durably.
Re-run the audit monthly; models, indexes and cited pages all change.

What does ChatGPT say about your company, and why does it matter?

ChatGPT says whatever its training data and its live web retrieval suggest is the consensus about you, presented with confidence. That description now reaches an enormous audience. OpenAI reported 900 million weekly active users and 50 million paying subscribers in February 2026, according to TechCrunch. Google said in May 2026 that AI Mode had passed one billion monthly users, per the Google I/O 2026 Search announcement.

When a prospect asks “is Acme legit” or “best payroll software for dental practices,” the assistant’s answer is often the first and only impression, with no second listing to click. If it omits you, misstates your pricing or repeats a two-year-old complaint, that is your position until the sources change. The good news is that assistants are consistent about where they look.

Which AI assistants should you audit?

Five, at minimum, because they read different parts of the web. Only an estimated 11% of domains are cited by both ChatGPT and Perplexity, and just 12% of URLs cited by AI tools overlap with Google’s top 10 organic results, according to 5WPR’s State of AI Citations 2026. A clean result in one assistant tells you little about the others.

ChatGPT (with browsing on): the largest consumer audience; leans heavily on Wikipedia.
Google Gemini and Google AI Mode: tied to Google’s index and Knowledge Graph; AI Overviews and AI Mode citations skew toward Google-owned properties, including YouTube.
Perplexity: citation-forward, with Reddit prominent among its top sources.
Claude: cites less often per response but pulls more sources when it does, per Muck Rack’s platform breakdown.

Muck Rack’s May 2026 analysis of more than 25 million links found ChatGPT included citations in 96% of responses (about five per response), Gemini in 82% (about eight) and Claude in 55% (about 13), according to Muck Rack’s “What is AI reading” report. Those citations are your map to the sources you need to fix.

Get your brand recommended by AIViral Spike Marketing audits what ChatGPT, Gemini and Perplexity say about you, fixes the sources behind the answers, and secures the coverage that moves them.Book a free consultation ›

The 12 prompts to run

Use a fresh session for each assistant, logged out where possible, with no memory or custom instructions. Replace the brackets. Record the full answer and every cited URL.

“What is [Company]?”
“Who is [Founder or CEO name] and what are they known for?”
“Is [Company] legit?”
“What do customers say about [Company]?”
“What are the most common complaints about [Company]?”
“How much does [Company] cost?” (or “What is [Company]’s pricing?”)
“What are the best [category] companies for [target customer]?”
“[Company] vs [Competitor A]: which is better?”
“What are alternatives to [Company]?”
“Has [Company] been involved in any lawsuits, scandals or controversies?”
“Where is [Company] based, when was it founded and who owns it?”
“Should I buy from [Company]?” (or “Should I hire [Company]?”)

Prompts 1, 2 and 11 test facts. Prompts 3, 4, 5, 10 and 12 test sentiment and risk. Prompts 7, 8 and 9 test whether you show up when no one names you, the commercial outcome that matters most. Prompt 6 catches stale or invented pricing.

How to score the answers

Score each answer on four criteria. Sixty answers (12 prompts times five assistants) fit on one spreadsheet.

Criterion	Question to ask	Score
Mentioned	Did the assistant name your company at all? (For best-of and comparison prompts.)	0 = absent, 1 = mentioned in passing, 2 = featured or recommended
Accurate	Are the facts (founding, leadership, location, products, pricing, ownership) correct and current?	0 = material error, 1 = minor error or outdated, 2 = fully correct
Favorable	Is the framing neutral to positive, with complaints in proportion?	0 = negative or warning tone, 1 = mixed, 2 = favorable
Cited sources	Which URLs are cited, and are they ones you can influence?	List every URL; tag each as owned, earned, review, database, forum or unknown

A sample row looks like this:

Assistant	Prompt	Mentioned	Accurate	Favorable	Cited sources
ChatGPT	7. Best [category] for [customer]	0	n/a	n/a	G2 category page, a 2024 listicle, Reddit thread
Perplexity	3. Is [Company] legit?	2	1 (old CEO named)	1	Trustpilot, Crunchbase, BBB, own About page

Then aggregate. Low “Mentioned” scores on prompts 7 to 9 mean a visibility problem. Low “Accurate” scores point to specific stale sources. Low “Favorable” scores usually trace to review platforms, forums or a news story. The cited-sources column tells you which to fix first.

Where do AI assistants get their facts about a company?

From third parties, mostly. The same Muck Rack study attributed 84% of AI citations to earned media, journalism accounted for 27% across more than 20,000 outlets, and paid or advertorial content for just 0.3%. 5WPR’s synthesis found Wikipedia held 47.9% of ChatGPT’s top-10 source share, Reddit roughly 46.7% of Perplexity’s, and brand websites largely absent from category top-10 citations in B2B SaaS.

The implication is uncomfortable for marketing teams: your own site is the least influential source in the set. Assistants trust what others say about you, especially Wikipedia, established news outlets, review platforms and forums. Paid or advertorial content, honestly, is barely cited at all; the value of any commissioned placement is credibility with buyers, consistent facts on authoritative domains, search presence and conversion. Independent journalism and accurate third-party records are what move AI answers. The mechanics are covered in how press coverage gets your brand recommended by ChatGPT.

How do you fix wrong information about your business in AI answers?

Work from the cited-sources column outward. Here is the correction workflow in the order that produces the most durable change.

Publish a canonical facts page and mark it up. One page with the exact company name, founding year, headquarters, leadership, ownership, products and pricing model, with Organization schema and sameAs links to every official profile. It will not be the top-cited source, but it is what editors, journalists and data providers check.
Correct Wikipedia and Wikidata, if eligible and only through proper channels. Request corrections on the talk page with independent sources and disclose your affiliation. If no article exists and you lack independent coverage, skip it; a promotional article will be deleted and the deletion debate becomes a new negative source.
Update Google Business Profile and the Knowledge Graph record. Verify the profile and fix categories, address and ownership. If a Knowledge Panel exists for the company or founder, claim it and suggest edits with supporting URLs, as explained in how to get a Google Knowledge Panel for a person.
Fix the databases. Crunchbase, LinkedIn, Bloomberg and Dun and Bradstreet listings, app stores, industry directories. Stale CEO names and old headquarters in AI answers usually trace to one of these.
Get the correct facts into press coverage. A staff-reported article, founder interview or trade feature stating current facts becomes a citable correction. Guaranteed editorial placements can help too, provided they are accurate; they put the right facts on a recognized, indexed domain even if rarely cited directly. Where negative coverage is the problem, see how positive press pushes down negative search results.
Work the review platforms. Respond on Google, Trustpilot, G2, Capterra, Yelp and the BBB, flag policy violations, and ask real customers for reviews through compliant channels. Assistants summarize complaint patterns from these sites almost verbatim.
Address forums honestly. If a Reddit thread is the cited source, a transparent reply from an identified company account plus a real fix changes what the next crawl sees. Astroturfing is detectable and makes it worse.
Use the platform feedback tools. Each assistant offers thumbs-down and report options, and Google’s AI features have feedback links. Use them for clear factual errors with the correct source, as a supplement; they rarely change an answer alone.
Publish original data or expert content worth citing. A benchmark, a pricing guide or a founder byline in an outlet with editorial standards gives assistants something accurate to quote.
Re-test monthly. Same 12 prompts, same five assistants, same scoring sheet. Track the four scores over time and note which sources appear or disappear.

How often should you check what AI says about your brand?

Monthly for the full audit, and immediately after any event that changes the facts: a rebrand, a leadership change, a funding round, a lawsuit or a wave of reviews. Retrieval indexes change daily and cited pages get edited, so a quarterly check misses too much.

Keep the audits in one file. The useful metrics are the share of best-of prompts where you are mentioned, the count of material factual errors, the ratio of favorable to negative framings and the number of cited sources you have influenced. Improvement shows in the cited-sources column first and in the scores a cycle or two later.

Because the sources it uses to build category lists include them and not you. Best-of answers draw on comparison articles, review-site category pages, community threads and news features. If competitors appear in independent “top 10” pieces and you appear only on your own site, the assistant has no third-party basis to include you. This is a visibility problem, not a ranking problem, which is why classic SEO wins do not carry over automatically; see why strong SEO no longer guarantees AI answer visibility. The fix is to earn inclusion in the sources assistants already trust for your category, and to make sure the facts they find there are current. A broader plan for that is outlined in AI search visibility services to get your brand recommended.

What Viral Spike does differently

Viral Spike Marketing runs this audit as the first step of any AI visibility or reputation engagement, using the same 12 prompts and scoring sheet, and hands the client the raw answers and citations. Correction work is then sequenced by cited source: facts page and schema, database cleanup, review response, and press coverage that states the correct facts in outlets such as Forbes, Newsweek and USA Today. Placements are guaranteed editorial features: you approve the draft, the article is published and indexed on Google, and the engagement is refunded in full if it never runs. We are direct that commissioned articles are rarely cited by assistants; their job is credibility and accurate entity data, while independent coverage and third-party records drive the answers. Details are on the reputation management agency page and the public relations service page, or contact the team to request an audit.

Run the audit this week

Open five fresh sessions, paste the 12 prompts, and score the 60 answers before you change anything. The scores will show whether your problem is accuracy, sentiment or visibility, and the cited URLs will tell you which sources to fix first. If you would rather have a team run the audit, sequence the corrections and secure the coverage that puts accurate facts in front of the models, contact Viral Spike Marketing and ask for an AI brand audit.

Frequently asked questions

Can I ask ChatGPT to correct information about my company?

You can submit feedback on a response and report factual errors, and OpenAI’s tools allow thumbs-down ratings with comments. This rarely changes future answers by itself, because the model draws on training data and live sources rather than a database of corrections. Durable fixes come from correcting the pages ChatGPT cites, such as Wikipedia, news coverage, review platforms and business databases.

Why does ChatGPT give different answers about my company each time?

Language models generate answers probabilistically, and assistants with browsing retrieve different pages on different runs. Personalization, memory and custom instructions also shift results. That is why an audit should use fresh, logged-out sessions, run each prompt more than once where possible, and focus on patterns across assistants and months rather than on any single answer.

Does my website matter for what AI says about my company?

Yes, but less than most companies expect. Research from 5WPR found brand sites largely absent from category top-10 citations in B2B SaaS, and Muck Rack attributed 84% of AI citations to earned media. Your site still matters as the canonical reference for facts and structured data that journalists, editors and data providers check, which indirectly shapes the third-party sources assistants prefer.

How long does it take to change what AI assistants say about a brand?

Source fixes on live-retrieval assistants such as Perplexity and Google AI Mode can show up soon after the corrected page is recrawled. Changes that depend on training data, or on new press coverage being published and indexed, take longer, often months. In our experience, plan for a quarter of sustained correction work before judging results, and track progress monthly.

Should I be worried if ChatGPT does not mention my company at all?

Yes, if your prospects use AI assistants to shortlist vendors. Absence from best-of and comparison answers means the third-party sources those answers rely on do not include you. It is fixable: earn inclusion in independent comparison articles, review-site categories and trade coverage, publish citable original content, and make sure your entity facts are consistent everywhere.