Grammarly’s Expert Review Needs Actual Experts

Grammarly just rolled out an "Expert Review" pane that pretends to channel guidance from Timnit Gebru, Kara Swisher, Casey Newton, and other recognizable voices. TechCrunch confirmed that none of those people are actually reviewing your drafts. The feature is simply rephrasing inference outputs while borrowing their names for credibility.

That is the kind of move that gets AI programs in trouble. When a writing copilot implies human endorsement, you inherit every disclosure, licensing, and misrepresentation risk tied to that persona. Imagine a regulated team shipping medical copy that cites a well-known clinician who never touched the text. That is how you end up with FTC consent decrees and contract clawbacks.

The fix is straightforward but non‑negotiable: force your copilots to cite real provenance, add automated checks that strip unauthorized names, and keep audit trails for every suggestion your agents surface. We build controller bots that sit between the AI output and the UI, flagging any hallucinated attribution before it reaches a customer or regulator.

If your marketing, compliance, or knowledge teams are experimenting with “expert” overlays, now is the time to harden them. We will help you add signature controls, inject accurate disclaimers, and prove to reviewers that no phantom endorsements are slipping through your stack.

Source: TechCrunch