AI Content at Scale: What Generated Pages Really Do in Google

A measured account of publishing generated pages at scale: coverage bought a five times rise in impressions and zero visitors, why average position decides everything, and the kill gates to write before you spend.

Short answer: coverage and traffic are different purchases. Adding generated pages reliably buys impressions, because Google will index and display almost anything coherent. It buys visitors only if those pages land somewhere a human actually looks, and on a weak domain they usually do not. Decide which of the two you are paying for before you generate a single page.

Featured Tools

Find adult performers now!

Hire Video Editors Directly

Adult Legal Compliance

Age-Verified Payments on Solana

100% Ethnic Erotic Porn

AI-Generated Latina Performers

Join Our Mailing List!Add your site?

Coverage buys impressions, not visitors

If my AI pages get indexed and show impressions, why is traffic flat?

Because impressions measure that Google was willing to show the page, and nothing else. On one content lane we ran, a coverage expansion produced a sustained rise in impressions from roughly 28 a day to roughly 156 a day, held over months, and produced no measurable increase in arrivals at all.

The reason is position. Those pages averaged a position in the forties. Almost nobody scrolls to the fourth or fifth page of results, so a page sitting there can accumulate thousands of impressions and never be clicked. The pages were not rejected by Google. They were seen and ignored by people, which looks identical in a coverage report and completely different in your revenue.

This matters because impressions are the first number to move and the easiest to celebrate. They move within weeks. Clicks, if they come at all, move much later. An operator watching the coverage graph gets a strong positive signal early and scales the lane, and the thing they scaled was the ability to be ignored more widely.

We had to formally withdraw an earlier internal claim on exactly this point. We had written that coverage is not a lever, and that was wrong: coverage moved impressions five times over and held. The honest version is that coverage bought reach and reach at position 41 is worth nothing. Quoting the clicks while dropping the impressions had made a real effect look like no effect.

Kill gates, written before the data exists

How do I know when to stop an AI content experiment?

You decide before you start, in writing, and you pick the threshold while you still have no idea which way it will go. A gate invented after the numbers arrive is not a gate, it is a rationalization, and everyone who has scaled a bad lane got there honestly by moving the goalposts one reasonable step at a time.

Here is the shape we now use on a cohort of a couple of hundred generated URLs. There are two gates, at two different times, measuring two different things.

Gate one, indexation, at six weeks. Inspect the URLs and count how many Google actually indexed. Below 30 percent the result is not a failure, it is inconclusive, and the correct response is to extend the window and look at crawl budget. Critically, below 30 percent you do not read the click data at all, because a click rate computed over pages that were never indexed is noise wearing a decimal point. Between 30 and 60 percent, judge only the indexed subset. Above 60 percent, proceed normally.

Gate two, impressions per page, at fifteen weeks. Not clicks. At that horizon the difference in clicks between a lane worth building and a lane worth killing sits inside normal noise, while the difference in impressions does not. Under 15 impressions per page per 28 days cancels the lane. Over 30 builds it out. In between means partial: ship the stronger half, not the whole thing.

One more check that is easy to forget and expensive to skip. Take a baseline of impressions on the parent pages before you stage anything, and compare it after. New pages can quietly cannibalize the pages that already earn, and a cohort that looks flat on its own numbers can still be costing you money elsewhere on the site.

Position decides everything, and the section decides position

Why do some of my generated pages earn clicks while identical ones earn nothing?

Because the template is not what is being judged. We ran two page families built from the same generator, on the same domain, with the same structure, and the click rate between them differed by roughly seventy times. Same demand, same code, wildly different outcome.

Inside one of those families the concentration was even starker. A single page took 100 percent of the family's clicks, sitting at position 8.5. Every other page in the same folder, built by the same template, earned zero: eighteen rows, 777 impressions, average position 56.9. One page was competitive and the rest were filler, and no amount of looking at the template would have told you which was which.

The lesson is that scale does not distribute quality. It distributes structure. Whether a given generated page earns anything is decided by whether that specific topic is one your domain can compete on, and that varies enormously between pages that look identical in your CMS.

The practical move: before generating a family, check where your domain already ranks on a sample of the exact topics the family will target. If the sample sits in the forties, generating two hundred more of them produces two hundred more pages in the forties. Fix position first, then scale. Scaling is the reward for having something that ranks, not the method for getting there.

X Reddit Bluesky Telegram