What if the misrepresentation is not a small factual error, but a systematic, negative or outright damaging portrayal of your brand by AI systems? Then a crisis protocol is needed.
When do we speak of a crisis?
- An AI system makes factually incorrect and damaging claims about your brand
- A competitor is systematically positioned as “better” based on demonstrably incorrect information
- An AI system fabricates statements or actions attributed to your brand
- The misrepresentation reaches a broad user group via a dominant platform (ChatGPT, Google AI Overviews)
Crisis protocol
Phase 1: Documentation (day 1-2)
Document all found misrepresentations fully: screenshot, date, platform, prompt phrasing, complete answer. This is your evidence file — essential for all subsequent steps.
Phase 2: Internal escalation (day 1-3)
Inform marketing, communications, compliance and legal. Determine the severity and potential impact. Establish a crisis team.
Phase 3: Direct correction of own sources (day 2-5)
Immediately update all own channels where incorrect information is or could be present. Publish a clear, factual correction page or statement that is directly indexable.
Phase 4: External outreach (day 3-10)
Contact the AI provider via official channels. Contact external publishers whose content serves as a source for the incorrect information. Activate PR contacts for corrective coverage in relevant media.
Phase 5: Monitoring and follow-up (week 2-8)
The timeframes in this protocol are internal response times, not expected recovery times: retrieval-driven errors can correct once sources are fetched again, parametric errors only at a model update — or not at all (see 9.3). Monitor intensively whether the corrections take effect, track on which channels the incorrect information still appears, and repeat the corrective measures where needed.
Prevention is the best strategy
A crisis protocol is necessary, but prevention is better. Organisations that structurally invest in GEO monitoring, correct external presence and systematic content updates are confronted with serious AI misrepresentation less quickly — and detect it faster when it does occur.