TrustCheck
Paste any AI-drafted email, proposal, blog post, or client memo and TrustCheck flags uncertain claims, made-up statistics, and weak logic so you can verify the draft before hitting send.
Independent consultants, freelancers, and solo agency owners who use ChatGPT or Claude to draft client-facing work
- Hallucination risk score (0-100) on any pasted AI output
- Source-claim verifier that highlights 'X said Y' statements to fact-check
- Inline auto-redline with 'verify before sending' notes
- Personal 'verified facts' memory so it learns what you've already validated
The viral repo shows AI agents routinely ship output that 'looks done but isn't checked' — that same failure is now showing up in freelance client work, where the solo professional takes the blame.
Real concern validated by lawyer sanction cases (Mazaheri v Law Society of Ontario) and a growing AI hallucination case tracker, but target segment of solo consultants checking drafts is narrow and most freelancers tolerate occasional errors.AI Hallucination Cases Database - Damien Charlotin ↗When AI Gets It Wrong: Lessons from Mazaheri v Law Society of Ontario ↗
Crowded fast: Veru, Exa hallucination detector (literally branded 'Grammarly for facts'), Olostep open-source detector, doany.app, and a 'Top 24 SaaS AI Fact-Checking Tools' roundup all already serve this exact paste-and-verify workflow.Veru | AI Citation Checker & Hallucination Detector ↗exa-labs/exa-hallucination-detector - GitHub ↗olostep-api/olostep-hallucination-detector - GitHub ↗
Hard to charge when the closest analogs (Veru, Exa, Olostep) are free or open-source; solo freelancers are notoriously price-sensitive, and listicle roundups show fact-checking tools competing on free tiers.Veru | AI Citation Checker & Hallucination Detector ↗exa-labs/exa-hallucination-detector - GitHub ↗
Hallucinations are likely a persistent LLM property so the problem endures, but OpenAI/Anthropic are increasingly baking grounding and citation into chat UIs, threatening standalone wrappers within 1-2 model cycles.The Top 24 SaaS AI Fact-Checking Tools in 2025 ↗
Buildable as a thin wrapper over search APIs (Exa, Olostep pattern), but every check costs real web-query tokens and accuracy on 'weak logic' (vs. factual claims) is genuinely hard — a non-trivial NLP problem beyond simple retrieval.exa-labs/exa-hallucination-detector - GitHub ↗olostep-api/olostep-hallucination-detector - GitHub ↗