When an AI assistant tells buyers your product is "buggy," "expensive" or "missing key integrations," it's usually not inventing it; it's summarizing something it read. The fix depends on whether the criticism is true, outdated or false. In every case, the way to change AI answers is to change the sources behind them: fix real problems, make fixes visible, update stale information and publish fresh, credible facts.
Step 1: Capture exactly what's being said
Run branded and comparison prompts across assistants several times and record the negative or outdated statements word for word, with platform, date and cited sources. Look for patterns: is it one phrase repeated everywhere, or scattered comments? See measuring brand sentiment in AI answers.
Step 2: Trace the sources
- Citations: read the pages cited in negative answers.
- Search the phrase: distinctive wording often comes from one review, article or thread.
- Review platforms: look for recurring themes in recent and older reviews.
- Forums and Reddit: search your brand with complaint words ("slow," "cancel," "support").
- News: past incidents, outages or controversies.
- Your own site: old pages that describe limitations or old pricing.
Step 3: Classify the claim
| Case | Example | Response |
|---|---|---|
| True and current | "Support response times are slow" (they are) | Fix the problem, then make the fix visible |
| True but outdated | "No mobile app" (you launched one last year) | Update sources and publish current information |
| False | "Owned by [another company]" (it isn't) | Correct the source; clarify on your site |
| Opinion or fit | "Too complex for small teams" | Consider positioning and content for that audience |
Responding to true, current criticism
There's no shortcut: fix the underlying issue. Then:
- Announce the fix in release notes, a blog post or a changelog, with dates.
- Update help docs and comparison pages to reflect the change.
- Respond on review platforms and forums where the complaint appears, as an identified representative, explaining what changed.
- Ask recent customers for reviews, so newer experiences appear. See reviews in AI recommendations.
Responding to outdated information
- Update or redirect your own old pages. See canonicals and duplicate content.
- Contact third-party publishers with a short, specific update request and a source.
- Update review platform profiles and respond to outdated reviews noting the change.
- Publish clear, current pages stating the facts plainly, such as a dated "What's new" page or an updated FAQ. See content freshness.
Responding to false information
- Correct the source with evidence.
- State the correct facts clearly on your entity home and in Organization schema. See organization schema and your entity home.
- Disambiguate if you're being confused with another company. See brand consistency.
- Use the AI product's feedback tools on clearly false answers.
- Seek legal advice for defamatory content, separately from AEO work.
Responding to opinions about fit
If AI says you're "too complex for small teams," ask whether that's fair. If it is, consider whether that segment is a priority. If it isn't fair, publish evidence for that audience: a small-team use-case page, a quick-start guide, case studies and reviews from small teams.
Handling a crisis
After an outage, breach or public controversy, AI answers may reflect it for a long time. Best practice:
- Communicate clearly and promptly on your own channels.
- Publish a factual summary and a resolution update when it's over.
- Keep that page live and dated, so AI can find your account of what happened and what changed.
- Avoid deleting or hiding information, which tends to backfire.
What not to do
- Post fake positive reviews or forum comments
- Argue with critics in public
- Flood the web with low-quality content to "push down" criticism
- Hide instructions to AI in your pages
- Threaten reviewers
These rarely work and often create the next negative story.
Track progress
Re-run the affected prompts weekly and record whether the negative or outdated statements still appear, and which sources are cited. Expect search-grounded answers to improve first, and model knowledge later.
The bottom line
Negative and outdated AI answers have sources. Find them, decide whether the claim is true, outdated or false, and respond accordingly: fix what's real, update what's stale, correct what's wrong, and publish clear, current information that AI can cite instead. Then track until the answers change.
