I wrote here yesterday about the one SEO rule I've used since 2004, and the part everyone asked about was the AI half. So here is the whole of it.
The site is a freelance marketplace I started in Japan in 2004. In March it did 14,000 sessions. Last month it did 90,000. The blog section on its own went from 2,300 to 52,000 over the same months, and ad revenue went up more than twenty times while sessions went up six.
The instruction is not "write an article about X". That gets you the page every other model produces, and Google has no reason to prefer it.
It is this. Take the term. Fetch the ten pages currently ranking on page one for it. Read all ten in full, not the snippets. List what they cover between them, and list what none of them covers. Write the piece that covers everything on the first list plus at least one thing from the second. Cite the source for every number. Then the run fails if the quality gate rejects it, and a failed piece is never published.
The gate is the part people skip. Mine checks that the claims have sources, that the piece isn't a restatement of one of the ten, that internal links exist, that the title uses the wording people actually type. A piece that fails goes nowhere, and roughly one in six does.
What this means in practice is that volume isn't the strategy, it's a consequence. Each page still has to beat ten specific pages to earn anything. Doing that by hand takes half a day. Doing it this way takes a few minutes, so 200 becomes possible, but the standard per page never moved.
And the policy question, since it always comes up: Google has never said you can't use AI. What gets caught under scaled content abuse is pages that exist mainly to rank. A page that beats the ten above it isn't in that category. A page that restates them is, however many you publish.
If you're doing something similar, what does your quality gate check?