Clay’s co-founder Varun Anand posted the company’s AI writing policy in August, on LinkedIn, under his own name. Not a memo. Not a wiki page. A public post.

Business Insider reported it drew nearly 10,000 reactions and more than 700 reposts. This week, Heidi Health’s Emma MacKenzie posted hers, per a recap from Outlever co-founder Melissa Rosenthal. Six principles this time, a different company.

Why a policy document gets 10,000 reactions

Nobody reads a policy for entertainment. What made Clay’s post travel is the company that wrote it.

Clay sells AI-powered outreach software. Heidi Health builds AI that drafts clinical notes. Both are in the business of generating text faster, and both just told the market, in public, exactly where that automation stops and a person’s judgment starts. That’s a harder thing to say than to write, and it reads as more credible than a policy from a company with nothing to lose by admitting AI writes badly on its own.

Ordinary readers already knew companies use AI to write. What they hadn’t seen was an AI company say it this plainly, without legal hedging.

What the two policies actually require

Clay’s four principles, written by engineer Sophie Alpert for the engineering team before the rest of the company asked for them: you own every idea in a document before you send it, writing is a form of thinking so skipping the draft leaves you with a weaker grip on the argument, a document should take longer to write than to read, and length is not a proxy for quality because AI happily pads a page with sentences that say nothing.

The line everyone quotes is the accountability test: if a reviewer asks what you meant by a sentence, “AI wrote that, ignore it” is not an answer you’re allowed to give.

Heidi Health’s version runs to six principles and, per Rosenthal, closes on a line with no self-congratulation in it: “we’re not the first company to do this, and we won’t be the last.” That admission matters. It reframes the policy as membership in a pattern, not a one-off flex. Polarsteps, the roughly 100-person Amsterdam travel app, and Leapsome, the 130-person HR platform based in Berlin and New York, published comparable policies of their own this year, each one requiring disclosure of AI use and final copy that reflects the writer’s own reasoning.

Four companies, four different products, in the same twelve months. None of them work in marketing. The people most exposed to a client asking, “Did AI write this?” are not the ones writing the policies yet.

What your marketing team’s version needs to cover

We deal with this daily at IQ Source, because the post you’re reading was drafted with Claude Code and runs through an AI-detection check before it publishes. That’s not a client story, it’s how we operate. And what that process taught us is that a policy missing any of these four points is a sentence with no owner behind it.

Who signs. Not “the content team.” One named person who answers if a client asks what a paragraph meant.

What AI can draft first, and what it can’t. A hard number, a client quote, or a product promise needs human authorship from the first line, not editing on top of a model’s guess.

When you disclose. Not every reader needs to know, but a client who asks directly earns a direct answer.

Who reviews before publishing, against what standard. Without that last check, the first three points are rules that exist only on paper.

We covered the same erosion from a different angle when Nature measured how much AI polishing strips out of a brand’s voice, and from the penalty side when LinkedIn started throttling reach for accounts that publish slop. The writing policy is the missing piece between those two problems: it decides, before anything gets published, who answers for what goes out.

If your company doesn’t have one in writing yet, the time to write it is before a client asks and there isn’t one.

Let’s build your team’s policy