Skip to content
moher.ai
← Journal6 min read

Publishing AI content without wrecking your rankings.

The difference between scale that works and scale that gets you deindexed is not the model. It's the gates.

Travis Moher

The debate about whether AI content ranks is the wrong debate. Search engines have never had a policy about how text was drafted; they have a policy about whether the page is useful and trustworthy. Plenty of AI-assisted content ranks. Plenty of human-written content deserves not to.

What actually distinguishes the two outcomes is the production system around the model — specifically, what has to be true before a page is allowed to publish.

Ground it in something real.

A model asked to write about a product it has no data on will produce fluent, plausible, unverifiable prose. That's the failure mode that gets sites penalised, and no amount of prompt engineering fixes it, because the problem is missing information rather than poor phrasing.

Retrieval against real sources — your catalogue, your test results, your live pricing — turns generation into assembly. That's a different and far more defensible activity.

Gate the claims that age.

Prices change. Products go out of stock. Statistics get superseded. A page that was accurate at publication and wrong six months later is a slow-acting trust problem, and at scale it becomes the dominant one.

Automated integrity checks — verify price against source, verify availability, retire anything that can't be confirmed — matter more than any single editorial decision. Retire rather than substitute: quietly swapping a dead recommendation for a different product is how a review site stops being one.

Keep a human on the hook.

Not a reviewer who rubber-stamps a queue. A named person accountable for the section, with the authority to reject and the time to use it. The gate only works if it sometimes stops something.

This is also the part that scales worst, which is exactly why it's the part that differentiates. If everyone could do it, it wouldn't be an advantage.

Structure for answer engines too.

A growing share of first impressions now happen inside an AI answer rather than on a results page. Getting cited there rewards the same things — specificity, verifiable claims, clean structured data — but punishes vagueness harder, because a summariser has nothing to summarise.

Write the sentence a machine could quote and a human would trust. That's the whole technique.

Related capability
Brand & content engines

Tell us what you're building.

Two honest paragraphs beat a ten-page RFP. We answer inside one business day, and the first call is 45 minutes with someone who can scope it.

Or write directly — t@moher.ai