AI Detectors Don't Work: The FTC Proof

Key takeaways

A US federal regulator, the FTC, found a popular AI detector (Workado's) performed "no better than a coin toss," with real-world accuracy of just 53 percent on general writing versus the 98 percent it advertised.
The core failure: the detector was trained only on academic writing, so it can't reliably judge the blog posts, emails, and sales pages you actually write. Its errors run both ways, flagging human writing as AI too.
The FTC order was unanimous (3-0) and forces the company to hold evidence for its accuracy claims and report compliance for four years. The lesson: confident accuracy numbers with nothing behind them are the problem.
Build the receipt before you need it. Draft in Google Docs or Word so version history saves itself, keep the original brief at the bottom of the same doc, and save the AI chat thread as your prompt-and-edit trail.
If a client accuses you: stay warm, ask which detector they used, send a view-only link and point them at Version history, and cite the FTC's coin-toss finding. Institutional cases run on their own rails, so lead with the evidence there.
Detector unreliability isn't a license to lie. Follow your contracts, check your facts, and own every sentence you ship. That ownership is what you can actually defend.

The Solo Creators AI Studio

Get this skill, and every one I build.

The whole vaultA new drop monthlyBuild live with meYours to keep, forever

See inside the Studio

30-day money-back guarantee. Cancel anytime.

Questions people ask

Can AI detectors really tell if my writing was made with AI?

Not reliably. The US Federal Trade Commission took action against Workado over its AI Content Detector, and the Bureau of Consumer Protection's director said it did "no better than a coin toss." Independent testing put its accuracy on general-purpose writing at 53 percent, versus the 98 percent it advertised. A tool that's right about half the time can't give you a trustworthy verdict.

Why do AI detectors flag writing that a human actually wrote?

Because they're trained on a narrow slice of text and then asked to judge everything. In the FTC case, the model "was only trained or fine-tuned to effectively classify academic content," yet it was sold to everyone. When a detector is roughly as accurate as a coin flip, its errors go both ways: it misses AI writing and it also accuses human writing. Clean, confident, well-organized prose gets flagged often, which is unfair to good writers.

How do I actually keep a draft history I can show a client?

Draft in Google Docs or Word instead of a plain text file, because both keep a timestamped version history automatically. In Google Docs it's File, then Version history. In Word it's File, then Info, then Version History for documents stored in OneDrive or SharePoint. Keep the original brief or outline at the bottom of the same document instead of deleting it, save the AI chat thread under the project's name, and if you ever need to prove your process, send a view-only link and tell the client to open Version history themselves.

What should I do if a client says my work was flagged as AI?

Stay calm and warm. First, ask which detector they used, since that tells you how much the verdict is worth. Second, send a view-only link to your working document and point them at Version history, which is real evidence a score can't match. Third, mention factually that the FTC found a popular detector performed "no better than a coin toss," and that these tools flag human writing as AI too. Redirect to what matters: whether the work is accurate and useful.

Should I use a "humanizer" tool to avoid getting flagged?

I wouldn't. A humanizer is a second unproven tool built to fool the first unproven tool, and it usually makes your writing worse by adding filler and roughness. You'd be paying two machines to argue about your voice while your actual reader gets a weaker piece. Write clearly for the human reading it, and keep your draft history if you ever need to prove your process.

Is it dishonest to use AI to write my content?

Using a capable tool well isn't cheating. The real questions are whether the work is true, whether you understand and stand behind every claim, and whether it solves the reader's problem. If a client's contract requires human-written work, follow it. Otherwise, drafting with AI and then owning and editing every sentence is a legitimate professional workflow. This isn't legal advice, so check any contract terms with your own attorney.

Did the FTC ban AI detectors?

No. The FTC didn't ban AI detection. It banned confident accuracy claims that aren't backed by "competent and reliable evidence." The unanimous 3-0 order lets the tools exist but stops the company from promising accuracy numbers it can't prove, and requires it to retain its evidence and report compliance for four years. The takeaway for you is to distrust any big accuracy percentage that has nothing behind it.

Found this useful? Share itPost Share

Written by Ronnie Nijmeh

Ronnie spent 18 years building a SaaS with a team of 20 that served over 650,000 customers, generated over $14M in sales, and sent over 550M emails. Now he's solo, solving real business bottlenecks and turning them into working AI skills, workflows, and automations. He teaches all of it, with direct access to him, inside the Solo Creators AI Studio Skool community. See what he's built →