AI keeps saying wrong things about our company. Has anyone successfully corrected misinformation in ChatGPT or Perplexity responses?

I specialize in exactly this problem. It’s fixable but requires systematic work.

Why AI gets it wrong:

ChatGPT’s training data has a cutoff date. If your old CEO was mentioned in 100 articles and your new CEO in 20, the model “remembers” the old one more strongly. It’s not malicious, just pattern matching on incomplete data.

The correction framework that works:

1. Source Audit List every place your company information appears online:

• Wikipedia (most important for AI)
• LinkedIn company page
• Crunchbase
• Official website About page
• Press releases
• Major news articles

2. Update Authority Sources These are the sources AI trusts most:

• Wikipedia - If you don’t have a page, consider if you’re notable enough. If you do, ensure it’s accurate with current citations
• Wikidata - Structured data that powers many AI systems
• Your website - Schema markup for Organization, including foundingDate, CEO, address
• LinkedIn - Company page with verified info
• Crunchbase - Frequently cited by AI

3. Fresh Content Strategy Publish new content that explicitly states correct information:

• “About Us” page with clear facts
• Press releases announcing leadership (even if not “new”)
• Blog posts that naturally mention current leadership, location, products

4. Report and Document

• Perplexity: Use their feedback button on each response
• ChatGPT: Report via feedback (yes, it feels pointless, but it’s logged)
• Document everything for potential escalation

Timeline expectation: 2-6 months for meaningful change. This isn’t a quick fix.

We went through this when ChatGPT kept saying we were an “e-commerce platform” when we’re actually a B2B SaaS tool. Completely different business model.

What finally worked:

The key was consistency across multiple authoritative sources.

We updated:

1. Website schema markup with correct industry, description, products
2. LinkedIn company page with detailed description
3. Crunchbase with correct categorization
4. Published 3 blog posts explicitly describing what we do
5. Got a TechCrunch mention that correctly described our business

Within 8 weeks, ChatGPT started describing us correctly. The TechCrunch article seemed to help most - authoritative tech publication + recent + correct information.

The lesson: AI trusts authoritative sources. Get correct info on high-authority sites, not just your own website.

Technical angle that people miss:

Schema markup is how you “tell” AI systems correct information.

On your About page, implement Organization schema with:

{
  "@type": "Organization",
  "name": "Your Company",
  "foundingDate": "2015-03-15",
  "address": {
    "addressLocality": "Austin",
    "addressRegion": "TX"
  },
  "employee": {
    "@type": "Person",
    "name": "Current CEO Name",
    "jobTitle": "CEO"
  }
}

AI systems increasingly parse schema markup to understand entities. It’s not a magic fix, but it’s one of the clearest signals you can send about accurate information.

Same for your products - use Product schema with current offerings only. Discontinued products should be removed from your sitemap entirely.

Important: You need ongoing monitoring, not just a one-time fix.

AI responses can change. We fixed misinformation for a client, and 4 months later ChatGPT started saying wrong things again because it ingested a new article with old information.

Set up:

1. Weekly manual checks of key prompts about your company
2. Am I Cited or similar tool for automated monitoring
3. Alerts for new web content mentioning your company (Google Alerts, Mention, etc.)

When you spot new misinformation, trace it to the source. Often it’s an old article that got republished or a new article citing outdated info.

This is an ongoing reputation management task, not a project with an end date.

What worked for us correcting product misinformation:

Create a “Company Facts” page on your website.

We made a simple page with:

• Founded: [Date]
• Headquarters: [City]
• Leadership: [Current names and titles]
• Products: [Current offerings only]
• Employees: [Approximate count]

Structured as a table with clear formatting. Schema markup for everything.

This single page became a reference that AI systems started citing. When you have one authoritative source of truth that’s clearly structured, AI prefers to cite it over piecing together information from multiple articles.

Legal perspective that might help:

If misinformation is causing actual business harm, document everything.

I’ve worked with companies pursuing formal complaints with AI providers. The key is demonstrating:

1. The information is factually incorrect (evidence of correct facts)
2. It’s causing measurable business harm (lost deals, confused partners)
3. You’ve attempted normal correction channels (report submissions)

For EU-based companies, GDPR gives some rights around accuracy of data about identifiable entities. It’s complex but has been used successfully.

For serious ongoing misinformation that normal channels don’t fix, some companies have had success with formal legal letters to AI providers. This escalates to a different team than the feedback forms.

However: Prevention is easier than cure. The source-update strategy others mentioned is more reliable than trying to force AI companies to correct responses.

Unexpected thing that helped us:

Get quoted in recent news articles with correct information.

AI systems (especially Perplexity) heavily weight recent content. We weren’t going to get a TechCrunch article, but we could:

• Submit to local business journals
• Get interviewed on industry podcasts (transcripts get indexed)
• Contribute guest posts to industry blogs
• Issue press releases through PR Newswire

Each of these created recent content with correct information that AI systems found and started citing.

Cost us maybe $2K in PR distribution over 3 months. AI responses started improving within 6 weeks.

Let me explain the technical reality of why this is hard to fix:

How LLMs learn “facts”:

ChatGPT doesn’t have a database of company facts it looks up. It learned patterns from training data. If your old CEO appeared in more training documents than your new CEO, the model “believes” the old info more strongly.

What this means practically:

1. You can’t directly “update” ChatGPT’s knowledge
2. You CAN update the web content that future training will use
3. You CAN influence real-time retrieval (ChatGPT’s browsing, Perplexity’s search)

For real-time retrieval (the part you can fix faster):

Perplexity searches the web live. If authoritative pages show correct info, Perplexity should cite correctly. Focus on making correct info the most prominent result for your company name.

For model knowledge (slower fix):

This changes when models are retrained on new data. OpenAI doesn’t announce training data updates, but they happen. Getting correct info onto authoritative sites now means future model versions will have better information.

Bottom line: Think of this as SEO for AI training data. You’re not fixing the model directly - you’re fixing what future models will learn.

This thread has been invaluable. Here’s my action plan:

Immediate (This Week):

1. Create Wikidata entry with correct structured data
2. Update website schema markup with Organization, CEO, products
3. Audit and update LinkedIn and Crunchbase profiles
4. Set up Am I Cited for ongoing monitoring

Short-term (Next 30 Days): 5. Create “Company Facts” page with clear structured information 6. Issue press release with current company information 7. Pitch local business publications for coverage 8. Report misinformation to Perplexity (with sources)

Ongoing: 9. Weekly monitoring of AI responses 10. Track improvements over time 11. Create new content that naturally includes correct facts 12. Consider legal options if significant misinformation persists

Key insight I’m taking away: This isn’t about changing AI directly - it’s about changing what AI learns from. Fix the sources, and AI eventually follows.

Will update this thread in a few months with results.