Official, observed, tested, hypothesis: how this site decides
Every sentence on this site that describes how a search engine works carries a status: official when the engine's vendor wrote it, observed when it can be seen in real data, tested when a repeatable experiment measured it, hypothesis when it is a reading this site owns up to. The status is displayed, and checked when the site is built.
Why label, instead of asserting?
Search is a field where almost nobody can check what they are told. The systems are closed, public data is thin, and the demand for clear answers is enormous. That combination mechanically produces confident writing: texts that explain, in one even tone, what has been documented for ten years and what was guessed last week. The reader has no way of sorting one from the other, short of finding the sources themselves — which nobody does.
Labelling costs writing time and rules out certain comfortable sentences. In exchange, it makes the text arguable. A claim whose origin is known can be contradicted, dated, revised; a bare claim can only be believed or rejected wholesale. It is also, very concretely, what makes a page safe for an answer engine to quote: a passage that separates the known from the assumed can be picked up without committing whoever picks it up.

What does each of the four statuses mean?
The four words are not degrees of confidence ranked from best to worst. They are four different natures of evidence, each with its own requirements, and the site cannot invent a fifth one to get itself out of a tight spot.
official
The engine's vendor wrote it or said it publicly, and the document is linked with the date it was consulted. It is the most demanding status, because it tolerates no generous paraphrase. One example, from a search engine's own documentation on sitemaps:
A sitemap helps a search engine discover URLs, but it commits to nothing about crawling them or indexing them.
officialSource : Google Search Central, “Learn about sitemaps”
The distinction looks slight and it changes everything: it forbids writing that a sitemap “gets a page indexed”, a phrasing that is everywhere. The engine itself says otherwise. You only have to read it.
observed
Seen in real data, with no proof of cause and effect. The status requires saying where the observation was made and on what. It never claims to explain: two curves moving together owe each other nothing, and a site that gains positions after a change may have gained them for a reason nobody saw.
tested
Measured by an experiment described precisely enough to be replayed: the question asked before looking at the data, the scope, what was changed and what served as a control, the measuring tool, the raw result even when it shows nothing, and a mandatory section on what the test does not prove.
The first three protocols were run on 16 August 2026, and all of them concern this site: does the content of a page depend on a script running, how many redirects separate each form of an address from the version actually served, and where the transferred weight goes. They are small tests, run on a single site, and that is written on them.
Two of them first produced a wrong result, corrected before publication: adding up every version of an image counted the same illustration several times, and a page's own address appearing inside its own markup was counted as one more resource. Both mistakes are described next to the correct result, because a protocol that hides its false starts cannot be replayed.
hypothesis
A reading the site owns up to, presented as such. The status requires no source, for a simple reason: it claims to prove nothing. It is the status of reasoning by analogy, of deductions from a partial document, and of everything that comes out of patents.
A patent filed by a search engine shows that an idea was protected, not that it runs in the service today.
hypothesis
Mistaking a patent for a description of the running service is the most widespread error in the field, and the easiest to make in good faith: the document is real, public, technical, signed. It proves exactly one thing, which is that an idea was protected.
How is an assessment card filled in?
A technique cannot be judged on a single axis. It demands an effort, it takes a certain time to have any effect, it exposes you to a risk, and its benefit lasts more or less long. Four questions, therefore four gauges, and the level always written out in words beside them — a colour alone never carries the information on this site.
Declaring a sitemap
hypothesis
- Effort
- low
- Time to effect
- weeks
- Risk
- none
- Durability
- lasting
This card carries the “hypothesis” status, deliberately. That a sitemap can be declared, and what it changes for an engine, are documented; saying that the effort is low, that the effect shows within weeks and that the benefit lasts is still an appraisal. An assessment card is a claim like any other: if it wants the “official” status, it has to point at a document that genuinely settles those four axes, and such a document is rare.
Three states exist alongside a normal assessment. Not assessed: the axes are drawn as dotted lines and no level is displayed, because there is no way to decide.Assessment disputed: the gauges are hatched, the sources do not agree, and the card says so rather than picking a side at random. Not advised: a thick frame, the risk axis at its maximum, and a mandatory warning on the same page — the card refuses to build if it is missing.
What does this site refuse to publish?
A site that explains a method has to say what that method excludes, otherwise the method is just a sales argument. Three refusals, held by the program that builds the site rather than by the writer's good intentions.
No promise of a position. Formulas like the following appear nowhere here, except in quotation marks so they can be taken apart:
First position on Google guaranteed in three months, results guaranteed or your money back.
No bare claim about a search engine. Two very common phrasings are refused by construction, because they present a reading as a fact:
Google loves long content, and content freshness is a ranking factor.
No monetising of risky techniques. The site may explain them, and it does — a reader who understands why a practice is dangerous defends themselves better than a reader who was told not to touch it. It may not earn a living from them.
Buying links to rank
officialGoogle Search Central, spam policies for Google web search
not advised
- Effort
- moderate
- Time to effect
- weeks
- Risk
- high
- Durability
- short-lived
Why are figures never written into a sentence?
A tool's price, a market share, a search volume: those values change without announcement and without history. Written into a sentence, they stay there for years and end up turning a useful guide into a false document. They all go through a separate display instead, which shows the date on which the value was checked and the source used to check it — the vendor's own page, never a third-party comparison.
Past a year, the value stops being displayed at all: the page states that the information has not been rechecked and sends you to the vendor. That is less satisfying than a figure, and far more honest than an out-of-date one. A reader checks a price in thirty seconds; the day they catch an error, they have no reason to believe the rest of the page.
The same reasoning applies to algorithm update timelines. This site publishes none: the lists going around are copied from one another, regularly wrong on dates, and they do not survive a check. An update is treated as a phenomenon observed in your own data, with a status and a protocol.
How does the site check itself?
None of the above would hold on personal discipline. A checking program reads every page before each deployment and stops publication at the first violation: a ranking promise, a claim about an engine with no status, an “official” status with no source, a “tested” status with no protocol, a commercial link with no visible disclosure, a risky technique with no warning, a figure in plain prose, an undated source, and a declared language version whose twin does not exist.
That last point deserves a word, because it shows the difference between following a specification and following a piece of documentation. Two things checked on the same day, in two different places:
The sitemaps.org 0.9 protocol defines no language-version element: the ones found in sitemaps are a namespaced extension.
officialSource : sitemaps.org, “Protocol”, version 0.9
In a sitemap, every URL that declares language versions must list all of those versions, including itself.
officialSource : Google Search Central, “Tell Google about localized versions of your page”
In other words, what many present as “the sitemap standard” is in fact a search engine's documentation, not the protocol specification. The confusion is harmless as long as everything works, and becomes a problem the day you look for the authority that settles the question. The robots exclusion protocol, by contrast, really is a specification:
The robots exclusion protocol is a specification published by the IETF, on the standards track.
officialSource : IETF, RFC 9309, “Robots Exclusion Protocol”, September 2022
What to do if a page on this site is wrong?
Reporting it is the most useful contribution anyone can make here, and it comes before other requests. A source that does not say what it is credited with saying, a status that is too generous, a date that has slipped: those are method errors, so they are the only ones that touch the reason this site exists. The contact address is in the legal notice, and the way this site makes money is described on the page devoted to it.
Frequently asked questions
- Why not simply write what everyone else writes?
- Because most of the claims circulating about search engines are readings presented as facts. Copying them would save time and cost the only thing this site has to sell: the ability to check what it says, line by line.
- Is a “hypothesis” status worth less than an “official” one?
- No, it is worth something different. A well-posed hypothesis often explains more than an official quotation taken out of context. What the status changes is the weight you give it: you do not make the same decision on a reading and on a document.
- What happens if a source disappears or changes its mind?
- The displayed date of consultation becomes the only useful landmark, and that is what it is there for. A page whose sources have aged is revised or unpublished. A document rewritten without a trace is common: that is precisely why nothing is cited here without a date.
- Does this site publish its own tests?
- Yes, since 16 August 2026: three protocols, all run on this site, all described precisely enough for a reader to redo them — what a page contains with scripting switched off, how many redirects each form of an address costs, and how the transferred weight is distributed. They stay classed as not established, because a home-made test on a single site is not a source from the search engine vendor.
Sources
- Learn about sitemaps — Google Search Centralaccessed on
- Tell Google about localized versions of your page — Google Search Centralaccessed on
- Spam policies for Google web search — Google Search Centralaccessed on
- Protocol, version 0.9 — sitemaps.orgaccessed on
- RFC 9309, Robots Exclusion Protocol (September 2022) — IETFaccessed on