A few months ago we noticed something annoying: we had close to a thousand live, legitimate pages on our site that Google had simply never indexed. Not penalized, not deindexed, just never crawled and picked up in the first place, some sitting untouched for over half a year. That's the kind of problem that's invisible until you go looking for it, because the pages exist, they just don't count.
We're a small team building Greta.sh, an AI tool for building and shipping software products, and like most small teams, SEO was the thing that always lost to shipping product features. So instead of hiring for it, we built an automated pipeline using Claude plus Ubersuggest for keyword and competitor data, and let it run the parts of SEO that don't actually need a human: finding pages that should be indexed but weren't, generating and shipping 50+ new pages in the past month, and continuously optimizing for AEO — the newer, less-talked-about discipline of making sure answer engines like ChatGPT and Perplexity can actually find and cite your pages, not just Google.
The mechanics, roughly:
Crawl our own sitemap vs. Google's index to find the delta (pages live but not indexed)
Feed that list + Ubersuggest keyword/competitor data to Claude to prioritize and rewrite/strengthen weak pages
Auto-submit and monitor indexing status instead of manually checking Search Console
Repeat weekly instead of "whenever someone remembers to"
Results after about a month: roughly 1,000 previously unindexed pages (some 6+ months old) got indexed, we shipped 50+ new pages, and — separately, but worth mentioning — redirecting our old questera.ai traffic into greta.sh gave us a 22% traffic bump on its own, which had nothing to do with content and everything to do with just not splitting our authority across two domains.
Top comments (1)
nice catch separating AEO from Google indexing, most teams still treat the delta crawl as the whole job. i use Viewfy for the citation side of AEO, and Ubersuggest-style keyword data still has to feed the rewrite step separately. worth pulling how many of those 1,000 pages actually get named by Claude or Perplexity, not just crawled by them.