Page discoveredGoogle can access it
PassSEO services
Find out why Google is ignoring you.
Most SEO is sold as a checklist. We start from how Google is documented and testified to work, find which part is failing on your site, and put the work in the order that lets it register.
GetBorvis / Organic visibilitySearch opportunity Sample view
04gates checked
IndexationPass
Intent matchReview
AuthorityGap
TechnicalContentAuthorityAI
The short version
Google does not read your site. It collects testimony about it.
Google scores relevance first, then gates on quality judged across the whole site, then re-ranks on thirteen months of what real searchers did, then re-ranks again with models trained on that behaviour. AI Overviews sit on top of the same index.
So the job is not to optimise a page. It is to make the site something searchers choose, people mention, and that holds facts existing nowhere else.
Customer searchesbest commercial cleaning company↗
Intent matchedThe page answers the wrong query
WeakResult chosenCompetitors carry stronger proof
LostThe real problemMore content will not fix the wrong page.
What is actually knownFact Leak Fact
Three things Google denied, then documented.
Clicks are a ranking signal
Google's VP of Search Pandu Nayak described the click re-ranking system under oath as "one of the important signals that we have". It runs on a rolling thirteen-month window, split by country and device.
US Department of Justice antitrust trial recordSite authority and a sandbox exist
The 2024 Content Warehouse leak revealed attributes named siteAuthority, contentEffort and hostAge, the last described as sandboxing fresh sites at serving time. The attributes exist. Their weights are unknown, and anyone quoting one is guessing.
Indexing is a quality verdict
On Search Off the Record, John Mueller said that "if our systems are seriously worried about the quality of a website, they will reduce the number of pages that they index", naming undifferentiated AI writing as an example: "Anyone could have written this. This tells me nothing."
Google Search Off the Record, July 2026We label every claim we make. Fact is Google documentation or court record. Leak means an attribute is known to exist but its weight is not. Study is a correlation someone published. Ours is our own inference, and we say so.
Six classes of witness01 02 03 04 05 06
The audit's first job is finding which one is at zero.
Each family is a class of witness testifying about your site. Adding more of a signal you already have produces less each time. A family sitting at zero is a gap nothing else fills, which is why sites with good links and good content still go nowhere.
Relevance
Term matching and embeddings, passage ranking, how well the title matches the query. This is the layer every provider works on, which is why competing there is expensive.
What we measure
Does the format of the page match the format Google is already showing for that query, and is the answer in the first screen?
Links
PageRank is alive and feeds site quality. Value depends on whether the linking page itself ranks, whether the anchor matches the topic, and how fast anchors appear.
What we measure
What share of pages linking to you rank for anything themselves? A third-party authority score answers a different question.
Clicks
Google keeps thirteen months of click history per query and per document, sliced by country and device. The strongest recorded signal is the click that ended the search.
What we measure
Is each page earning the clicks its position should earn? A page under-performing its position is testifying against itself.
Quality
Judged across the whole site, and partly at indexing time. Thin sections decide what the strong sections are allowed to do.
What we measure
How many indexed pages have earned no impression in twelve months? That ratio is the fastest read on the gate.
Entity and brand
Whether the business is a recognised thing with consistent facts, and whether people search for it by name.
What we measure
Is branded search growing over sixteen months? It is the leading indicator for both classic and AI visibility.
Demotions
Scaled content, borrowed reputation, doorway pages, anchor mismatch. These are not deductions. They are multipliers applied after everything else.
What we measure
Is anything on the site capped? One demotion cancels every positive above it.
Order of operations
Positives add. Demotions multiply.
This is the reason behind the most common complaint in this industry, which is that a site had work done and nothing moved. If a demotion is capping the site, every positive added underneath it is multiplied by something close to zero.
So the sequence is fixed. Remove what is capping the site. Fill the witness class sitting at zero. Then grow. Buying links or publishing more on a capped site sells work that cannot register.
Ours The multiplier model is our reading of the evidence, not a Google statement. We hold it as a working hypothesis and we say when it fails.
Honest timelines
Google runs on four clocks. Ours match them.
Nobody can promise you a ranking, and we do not. What we can do is tell you which clock a piece of work is on, because that decides when you could reasonably see anything at all.
| Work | Clock | Earliest reasonable signal |
|---|---|---|
| Titles, snippets, freshness on pages already eligible | Fast | 4 to 6 weeks, measured as click-through against the curve for the position held |
| Relevance, page structure, answering the sub-questions | Model retraining | Roughly 70 days, because the relevance models are retrained on a cycle |
| Click history, branded demand | 13 months | The click record is a rolling year. A bad title left up is bad testimony remembered |
| Anything touching site-level quality | Slow | Usually months. Google says improvements do not have to wait for a major core update, because it also runs smaller unannounced ones. If nothing has moved after a few months, the next broad core update is the next likely point |
If a site has been hit by a site-level classifier, recovery is slow, sometimes partial, and sometimes does not arrive. We say that before money changes hands, not after.
Ours These clocks are our working estimates from Google's statements and trial testimony, not schedules Google publishes. We corrected the last row on 10 October 2026: it used to say site-level changes only register at core updates, which Google's own guidance contradicts.
We run it on ourselves first
Our own site failed part of this audit.
On 23 September 2026 we measured every rendered page on this site. Eleven of thirty were under 600 words. Two of those were these service pages, at 334 and 367 words, thinner than every research piece we had published.
We had been writing essays and leaving the shopfront empty. This page is the fix, and we are telling you because an agency that cannot find its own problems is unlikely to find yours.
Read what we publish ↗SEO and AI visibility
If Google cannot use the page, its AI features cannot either.
AI Overviews and AI Mode run on the same crawl, the same index and the same ranking. A page has to be indexed and eligible for a normal snippet before it can support an AI answer.
"You don't need to create new machine readable files, AI text files, or markup to appear in these features. There's also no special schema.org structured data that you need to add."
Fact Google Search Central, AI features documentation
We publish an llms.txt file and we would tell you not to buy one. It is documentation hygiene, not a tactic.
That makes SEO part of AI visibility rather than the whole of it. ChatGPT, Perplexity and Claude retrieve differently, and each has to be checked on its own terms.
See how AI visibility works ↗Find the reason you are not ranking.
Free first check. We name the one thing most likely holding the site back, and the clock it sits on. No sales call. No automated score dump.
Request a website check ↗Agency rebuilding a client’s site? See the launch handoff, from $1,200. Building something else? See how we partner.
Sources
- Pandu Nayak's testimony on how Search ranking works, US Department of Justice antitrust trial.
- Analysis of the Google Content Warehouse API leak, May 2024. Attribute names are documented; weights are not.
- John Mueller and Martin Splitt on indexing and site quality, Search Off the Record, July 2026.
- Google's own guidance on AI features in Search, which states no special markup or files are required.
Figures and platform behaviour described here were checked on 23 September 2026. These systems change, and a dated claim describes its date. If something here turns out to be wrong, we correct it on the page with the date attached.