ChatGPT hallucinations can damage your SaaS business

Executive overview

ChatGPT began recommending geocoding API OpenCage for a phone lookup service they have never offered. Dozens of angry signups flooded in daily, unable to understand why the API was "broken." The problem is AI chat tools confidently fabricate details — services, code, author credits — with no built-in signal of reliability.

Treat AI-generated recommendations about your product as a potential liability, not free marketing.

How OpenCage was harmed

• Signups citing ChatGPT as referral source started arriving in January 2023
• New users exhausted free-trial API calls, then stopped — the feature they wanted did not exist
• ChatGPT generated working-looking code to hit a non-existent API endpoint
• Root cause: old YouTube videos incorrectly described OpenCage as offering phone lookup; ChatGPT absorbed that false data
• Phone location lookup from a number alone requires carrier-level access — impossible for a third-party API to build
• OpenCage published a blog post titled "Don't Believe ChatGPT — We Do Not Offer a Phone Lookup Service"

Why this happens

• AI models absorb false information from the internet and reproduce it with high confidence
• Users have learned to be sceptical of human-written content but treat AI output as authoritative
• There is no feedback channel: you cannot tell OpenAI to correct a single result in their training corpus

What to do about it

• Take all AI-generated content with a grain of salt, especially for product or API capabilities
• Demand references: AI outputs that surface their sources (like Google snippets with URLs) can be evaluated for credibility
• Monitor your own brand in AI outputs — hallucinations about your product affect support load and reputation