Programmatic location pages made with AI: how many before they look thin?

Say a dental clinic wants a page for every neighborhood it serves. An AI assistant can draft forty of them in an afternoon, each with the area name, a parking note and a short FAQ. The risk is that when only the town name changes, Google tends to crawl the pages and then leave them in the not-indexed pile, or they pull almost no impressions. Some people say the fix is just more unique wording per page, and I'm not sure that holds, because unique sentences that still say the same thing can look just as thin. I'd start with six or eight pages that each have a real reason to exist, like a local case, a service offered only at that branch, or photos from that office, and see whether they index before scaling. What's the smallest set of location pages you've seen actually earn impressions?

Starting small makes sense, but indexing alone doesn't tell you much. In Search Console, filter performance by page and look at which queries those pages get impressions for. If a neighborhood page only shows up for the clinic's brand name, the neighborhood part isn't doing anything yet. Then tag the call and booking clicks on each page in GTM with a location value, so you can see which pages actually produce enquiries and not just crawls. That's the signal for whether the other thirty-odd pages are worth making.

I wouldn't use per-page enquiries as the go/no-go signal. A single clinic's neighborhood page might get two or three calls a month, so one extra call can swing the numbers and you'll never see a real pattern. On top of that, the GTM tag only fires on clicks you can measure. If the clinic's phone number isn't a tracked tel link or a call-tracking number, most calls from a phone search never show up at all.

I'd judge the set instead of each page. Launch the six or eight pages, then compare their call and booking rate against the clinic's other pages over two or three months. What you really want to know is whether patients from those neighborhoods show up at all. If the booking form asks for a postcode or neighborhood, that answer is a cleaner check than per-page clicks.

I'm not sure comparing against the clinic's other pages holds up. Service pages and the homepage draw people with different intent, so a higher call rate there doesn't tell you much about the neighborhood pages.

A cleaner test is to build pages for only half the neighborhoods you plan to cover and leave the other half without one for a couple of months. Say a clinic serves eight areas and builds four pages first. Then compare booking share from those four areas against the other four, using the postcode field on the form. You still need roughly similar search demand on both sides, but that comparison is much easier to read than a page-by-page scorecard.

Agreed on the split. Before you pick which four areas go first, pull a geographic report for your core keywords in Microsoft Advertising. It breaks impressions and clicks down by city and metro area, so you can confirm the two groups have similar search demand instead of guessing. Otherwise a gap in bookings may just reflect a gap in searches. And if you track indexing, Bing Webmaster Tools has its own URL inspection, so a page can be indexed in Bing and still be missing from Search Console, or the other way round.