Google was indexing 9 of this practice's 67 pages. Six weeks after the fix, clicks had grown 2.4x.
Dr Arnaud Vali runs a dental practice in Neuilly-sur-Seine with an implantology focus and a well-built website that Google was ignoring: 49 of its URLs had never been crawled at all. The problem was not the content, not the links, and not the canonicals. This is what diagnosing and fixing crawl starvation looks like, measured in Search Console.
A good site that Google refused to read
The practice had invested in a proper Webflow site with real content. Search Console told another story: 9 of 67 pages indexed, and 49 URLs marked "discovered, currently not indexed" whose last crawl date read January 1, 1970, meaning never. The money pages for implants, aesthetics and prosthetics simply did not exist for Google.
Around thirty manual indexing requests over six months had changed nothing. A wave of twelve new articles published in April fell into the same void.
Not a penalty. Starvation.
We eliminated the usual suspects one by one, with evidence: robots.txt, noindex, canonicals (every page canonicalized correctly), Cloudflare blocking, soft 404s, duplicate content, thin content, manual actions. All negative.
The crawl stats told the real story: Google allocated about 4 requests a day to the domain, wasted 44% of them on redirects and 404s inherited from the domain's previous life, and spent only 4.4% of its crawl on discovering new URLs. The site was not penalized. It was starved.
Two facts that killed the easy explanations
Four moves, dated July 15-16, 2026
Stop wasting the crawl Google was willing to give
Dead URLs removed, seven precise 301 redirects, a clean sitemap, and internal navigation that exposes the service hubs instead of hiding them.
Rebuild the implant cluster as one machine
27 pages published in two days: the pillar rewritten from 4 sections to 9 sections and 31 subsections at 2,518 words, six sub-pages, ten converted articles and ten new pages on low-competition queries.
Make every page machine-readable
Structured data went from 1 page to 27: practice entity, the practitioner as author and reviewer, breadcrumbs, and medical page plus FAQ markup tied to the visible content.
Measure it like an experiment
Same grid points, same queries, before and after. And one control variable: the review count did not move during the window, so the growth below is search work, isolated from reputation effects.
Six weeks after the rebuild, measured
Sources: Search Console API extraction of August 28, 2026, and Local Falcon 25-point grid scans on identical points and queries. Nothing estimated by hand.
| Metric | Before | After | Change |
|---|---|---|---|
| Clicks (GSC) | 72 in the prior two months | 173 from Jun 22 to Aug 25 | 2.4x |
| Impressions (GSC) | 5,910 | 14,029 | 2.4x |
| Weekly impressions | about 700 | 2,400 to 2,800 | 3.5x |
| Pages receiving impressions | 15 in June | 27 in August | +12 |
| Map rank at the address, "dentiste" | #4 on Jun 19 | #1 on Jul 13, still #1 on Aug 28 | #1 held 6 weeks |
| Grid points in top 20, "cabinet dentaire" | 1 of 25 | 13 of 25 | 13x |
The map grid on "cabinet dentaire", June 19 vs August 28
Each point is a real ranking measurement around the practice. In June, 1 point of 25 was in the top 20. In August, 13 of 25.
The control variable: zero new reviews
Between June and August the practice's review count stayed exactly where it was. Everything above happened without a single additional review, which is what lets us attribute it to the technical and content work rather than to reputation. It also tells us the next lever: the review gap with local competitors is now the limiting factor, and we said so to the client instead of selling more content.
While SEO compounded, ads bought the short term
Implant campaign, June 27 to August 10, 2026: 629 clicks at a 7.87% click-through rate on 609 EUR of spend, 16 tracked calls, and about 44 EUR per identified patient contact. Paused on August 10 once the organic engine was producing.
No client pays Forge to be featured
Verifiable at source
Every figure comes from the Search Console API, the Google Ads console or a dated grid scan. We keep the raw extractions and will screen-share them with the client's permission.
The ugly baseline is on the page
Nine indexed pages out of 67 and thirty ignored indexing requests is not a flattering starting point. It is the real one.
Named, not anonymous
A real practitioner, a real practice in Neuilly-sur-Seine, a real website you can search for right now.
About this case
Why was Google only indexing 9 of 67 pages?
Crawl starvation. The domain received about 4 crawl requests a day, wasted 44% of them on redirects and 404s inherited from its previous history, and spent 4.4% on discovery. Robots.txt, noindex, canonicals, duplicate content and penalties were all ruled out with evidence first.
How fast did results come after the fix?
The rebuild shipped July 15-16, 2026. The weekly impressions curve inflects that same week, and over the following six weeks clicks ran at 2.4 times the prior two months, with pages receiving impressions going from 15 to 27.
What did the local map work produce?
On identical 25-point grid scans, the practice went from #4 to #1 at its address on "dentiste" between June 19 and July 13, 2026, and held #1 through August 28. On "cabinet dentaire", top-20 grid points went from 1 of 25 to 13 of 25.
Is this repeatable for another practice with indexing problems?
The diagnostic method transfers: rule out the ten usual suspects with evidence, read the crawl stats, then stop wasting crawl and give Google a dense, structured cluster worth crawling. The exact numbers depend on the domain's history and market.
Find out what Google is ignoring on your site
On the call we pull your Search Console live, read the crawl stats the way we did here, and show you exactly which pages Google has never seen.
Book a call and walk the dashboards live