The process
How we fix AI errors — systematically.
A rigorous, four-step methodology built to work within the official correction processes of every major AI provider.
Step one
01 Audit
We run structured queries across ChatGPT, Claude, Gemini, Copilot, and other major AI models to capture exactly what they say about you. Our audit covers key facts: biographical details, publications, products, statements, affiliations, and historical records. Every response is logged, timestamped, and cross-referenced against verified sources.
- Multi-model coverage across all major AI systems
- Structured query methodology for consistent results
- Cross-referenced against authoritative sources
- Severity scoring for each finding
Step two
02 Identify
Not all errors are equal. Our analysts classify each finding by type — fabrication, outdated information, misattribution, or omission — and assign a severity score based on reputational impact. You receive a comprehensive report with every error documented, ranked, and contextualised.
- Error classification by type and severity
- Reputational impact assessment
- Comprehensive written report
- Priority ranking for correction effort
Step three
03 Correct
Working with you, we build structured correction datasets that meet each AI provider's specific requirements. This includes sourced factual statements, verification documents, and contextual metadata. We handle all the technical formatting — you provide the authoritative truth.
- Provider-specific dataset formatting
- Source documentation and verification
- Collaborative review with your team
- Version-controlled correction records
Step four
04 Submit
We submit correction packages through each provider's official channels and track progress. Where providers have public correction mechanisms, we use them. Where they don't, we engage through established research and enterprise channels. Every submission is tracked until resolution.
- Official channel submission to each provider
- Progress tracking and status updates
- Follow-up and escalation where needed
- Resolution confirmation report
Common questions
How long does the process take?
An audit typically takes 5–7 business days. The correction and submission phase depends on the number of errors and AI provider timelines, which can range from 30 to 90 days.
Can you guarantee corrections will be accepted?
We can't guarantee acceptance — AI providers make their own decisions. What we can guarantee is that your corrections will be submitted correctly, through the right channels, with the strongest possible documentation.
Which AI models do you cover?
We currently cover ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), Copilot (Microsoft), and Perplexity. We're adding new models regularly.