SEO & organic search

Advertising buys a slot: the advertiser chooses the queries they want to appear on, pays when someone clicks, and the ad shows as soon as the campaign is live then disappears when it stops. Search optimisation buys nothing; no sum paid to an engine places a page there, and selection depends on the engine's estimate of how useful that page is for the query. Hence three differences in kind: advertising is immediate and reversible, optimisation slow and cumulative; advertising is finely controlled, optimisation is subject to an interpretation; advertising ends with the budget, the underlying work keeps producing an effect. Running both is coherent, provided neither is judged by the criteria of the other.

The honest answer is that no timescale can be announced without lying, and that anyone announcing one should be treated with suspicion. The spread depends on factors that do not offset one another: the technical state of the site and the pace at which an engine recrawls modified pages, the history of the domain, competition on the targeted queries, the amount of content to produce, and how fast the subject itself evolves. What can be observed instead is an order: indexing corrections show before content effects, which show before reputation effects. Following that order lets you verify that the work is progressing even while ranking has not moved yet.

No, and the commitment itself is a warning sign. A position does not belong to the site: it results from a comparison made by a third party, for every query, on unpublished criteria that change, against competitors who are working too. Nobody can promise what they do not control. Add that a position is no longer a single value: it varies with the device, the language and the context of the query, and the place of the first conventional link depends on what the engine displays above it. What can reasonably be committed to concerns the work: what will be corrected, what will be produced, what will be measured, and by what method the following decisions will be taken.

Two complementary means exist. The first is to query the engine itself for the pages it knows about a given site, which gives a quick but approximate indication. The second, more reliable, goes through the tool engines make available to site owners after ownership verification: it states page by page whether the address is indexed, and if not why — page discovered but not crawled, crawled then set aside, judged equivalent to another, or excluded by an instruction. That distinction is the starting point of any diagnosis: as long as a page is not indexed, no question of ranking arises and reworking its content will change nothing.

Rarely, and the finding is counter-intuitive. A heavily searched query gathers people with very varied intents, many of whom are not looking for a supplier; it is moreover occupied by established sites. A more precise query gathers fewer people but people who know what they want, and the difference in competition makes it reachable. The reasoning to hold is therefore not "how many people phrase this query" but "among them, which are in a position to decide, and am I able to answer better than what is already displayed". A good clue sits in the existing results: if they bear no resemblance to what you offer, the engine has understood that query differently from you.

Regularity helps, but it is not the objective; what counts is covering the intents that matter, not the publishing rhythm. Publishing for the sake of publishing produces weak pages that compete with each other and dilute the subject of the site. Two reflexes are worth more: improving an existing page that already appears without getting clicks, which starts from a real signal; and creating a page only when an identified intent does not yet have one. Updating also has a value of its own, notably towards systems that compose answers and must choose between contradictory sources: a page kept current and dated is selected more readily than one that gives no indication of whether it still holds.

Three requirements cover the essentials. Each version has its own address, stable and legible, rather than a display conditioned by a browser setting. Each page declares its equivalents in the other languages, so the engine understands that these are versions and not copies, and offers the right one. Each version is written in the target language rather than mechanically translated, because people do not use the turns of phrase that come out of a literal translation and selection turns precisely on those words. A practical consequence follows: the structure can be shared, the phrasing is not, and it is normal for one version not to have exactly the same pages as another if the queries differ.

No, and the risk is asymmetric. A link weighs because it is supposed to express a recommendation: someone found the page useful enough to send their readers to it. Paying to obtain it removes what gave it its value, and engines devote significant resources to spotting such arrangements. The outcomes range from ineffectiveness — the link is ignored, the spending is lost — to a penalty, by way of the devaluation of other legitimate links from the same site. A rented link moreover disappears as soon as you stop paying. What produces durable links asks for more: publishing something that deserves citing, taking part where the subject is discussed, and becoming visible other than through your own site.

They move the selected unit: a system that composes an answer retains not a page but a passage, retrieved after breaking the question down into several searches. Three consequences follow. A paragraph that stands on its own, that states before it qualifies and that defines the terms it uses stands a better chance of being taken up than a development dependent on its context. An attributable page — explicit subject, dated update — is selected more readily against a contradictory source. And visibility stops being measured in visits alone, since one can be cited without being consulted. The rest does not change: there has to be a page, a crawler must be able to read it, and there must be a reason to prefer it to another.

By measuring in the order in which things happen. First indexing, without which nothing follows. Then appearances, which say for which queries the engine judges the page relevant and therefore whether it has understood its subject. Then clicks, which depend on the title and summary displayed as much as on the position. Finally the enquiries obtained, the only measure that interests a business, provided you know through which page and which query the person arrived. Two precautions make the reading reliable: look by group of intents rather than as one global curve, and keep a dated log of changes, without which no effect can be attached to a cause.