That is the only question I answer, and I answer it with the working shown — including the parts that turned out to cost me money.
Ajay Singh · Paisley, Scotland · working with UK and international clients
A site with tens of thousands of pages goes quiet in search. Search Console shows a wall of “not indexed”, a validation that failed without giving a reason, and no obvious cause. Somebody says it was an algorithm update. Usually it wasn’t.
I take that apart and tell you what actually happened, in writing, with every figure reproducible. Then you decide what to do about it — with your own team, your own agency, or nobody at all.
Diagnosis of indexing and crawl problems on large sites. Why pages were dropped, which bucket they fell into and what that bucket actually means, whether the sitemap and the pages contradict each other, and what the numbers in Search Console are and are not telling you.
Content, link building, ongoing SEO management, or anything I would have to learn on your budget. If the answer is that you need one of those, I will say so and that is the end of my involvement.
A technical publisher with roughly 41,000 pages across two hosts. Search Console showed 38,600 not indexed and a failed validation. He believed a Google update had hit him.
It hadn’t. Google had indexed 2,818 of his 2,820 hub pages, held them about six weeks, then dropped them inside a fifteen-day window in July. Search Console holds no data across most of that window, so the fall appeared as a blank gap rather than a cliff — which is why nobody had found it.
Crawled – not indexed is a quality judgement about pages Google has read. Discovered – not indexed is a budget decision about your site: Google has queued those URLs and concluded they are not worth fetching. They sit next to each other in the same report and mean opposite things. Improving a page cannot help if Google will not come back to look at it.
The recommendation was fewer pages, not more. Rebuild the 2,820 hub pages so each carried its own answers, then consolidate the 34,400 thin ones into them.
Part-way through, the client sent his analytics. His data broke my recommendation. I had been about to advise redirecting all 34,400 pages. His figures showed Bing — which supplied 304 of his 334 real monthly search visitors — landed almost entirely on the pages I was proposing to redirect.
The fix was small: exempt the roughly 180 pages Bing actually uses and redirect the rest. Half a percent of the pages kept, and it still clears more than 99% of the crawl queue. That correction is written into the deliverable rather than quietly folded in.
Without it I would have been recommending a version of exactly the outcome I had just argued against — and it was your own data that caught it, not mine.
2,196 of his monthly visitors arrived as “direct”. I believe they were automated: there is no route to those URLs, the daily shape is wrong for humans, and each visit fetched exactly one page on a template carrying 33 internal links. That is inference, not measurement, and the document says so. The check that would settle it is a day of server logs grouped by user agent, which only he can run.
Three arguments pointing the same way is not proof, and a client is entitled to know which of the two they are being handed.
This is a diagnosis case study, not a results case study. The work was delivered in August 2026 and nothing has been implemented yet, so there are no ranking or traffic outcomes to claim and none are claimed here. When outcomes exist this gets a results section, and not before.
You add me to Search Console as a user. No passwords, ever. I work from that plus public data, and you get a written document with the method behind every number, an appendix showing how each was measured, and a plain list of what I could not establish.
Typical engagement £499. Larger or multi-property sites are quoted before anything starts, and I will tell you if I do not think there is a finding worth paying for.
Indexing problems are quiet. They do not announce themselves and they are usually found months late, which is most of what makes them expensive. Ongoing monitoring is a monthly arrangement, no contract, cancel whenever it stops being useful.
Short, no face, no intro. Real findings, with the check that proves each one.
Crawled vs Discovered — not indexed
The two lines in Search Console that mean opposite things, and the test that tells you which one you have.
Two of Britain's biggest retailers looked 100% broken to my own scanner. They weren't — a refused request isn't a fault, and most audit tools can't tell the difference.
One slash de-indexed a whole site
A canonical tag and a redirect disagreed by one character. Google kept neither URL until it was fixed.
Same channel, more of the same: youtube.com/@indexingdiagnosis
Send me the domain and one sentence about what looks wrong. I will tell you within a day or two whether I can see a fault worth paying to have explained — and if I cannot, I will say that instead, which happens often enough to be worth mentioning.