SaaS SEO

Most SaaS SEO plans start from a keyword export sorted by volume. That produces a content calendar, not a strategy, because volume tells you how many people search — not whether any of them buy your category, and not whether the query will ever produce a click at all.

The first stage separates three things that a volume column collapses into one: demand that converts, demand that exists but will be answered without a visit, and demand that looks large because it belongs to somebody else’s audience. All three are worth knowing about. Only the first one justifies a roadmap.

This is also where the honest no happens. If a category has no measurable demand, or the product does not convert the traffic it already has, that is the finding — and it is cheaper to learn it in week one than in month nine.

The most common state I find is a site with two hundred blog posts and no comparison page, no alternatives page, no integrations pages and no use-case pages. The blog is doing top-of-funnel work while the layer a buyer actually reads while choosing a tool does not exist.

That layer gets built first, and deliberately, because it converts while the content layer is still compounding. A comparison page ranks against a handful of competitors rather than the entire internet, and the person reading it has already decided to buy something.

The shape matters as much as the existence. One comprehensive page per commercial question beats twelve thin variations of it — for ranking, and even more so for being quoted.

Content at this stage is not volume. It is the specific job of making a page the most complete, most extractable answer to the question it targets — which is now the same requirement for ranking and for being cited.

The two choices that do most of the work are unglamorous: put the answer above the fold, and mark up real data as real data. A number, table or list in the first screen produces both the click-through rate and the citation. A table rendered as an image is invisible to a crawler and useless to a model.

The third choice is the one nothing can copy: first-hand detail. On a SaaS site that is the thing you know because you run the product — the limits, the edge cases, the migration that goes wrong, the real pricing behaviour. It is also the part a competitor generating pages cannot fake.

Authority work is usually sold as link building. Links matter, but the thing being built is narrower and more useful: becoming a source that gets referenced on a subject, rather than a site that ranks for a keyword.

Two moves do most of it. The first is covering the adjacent questions, not just the commercial ones — the explainers nobody is bidding for, which cost little competitively and establish the site as a source across the whole subject rather than only at its shopping end.

The second is a named, credentialled author on the pages where trust is the deciding factor. An answer engine that can attribute an explanation to a person will attribute it; a page with no author is a page with nothing to attribute.

Buyers increasingly get a shortlist from an assistant before they visit any website. If the product is not in the answer, the click never happens — and no amount of ranking beneath the answer fixes it.

This is not a separate discipline bolted onto SEO, and it does not involve special markup or files — Google states none are required for its AI features. It is the same entity clarity, structure and authority work pointed at a surface that quotes instead of linking, which is why it sits at stage five rather than being sold as a product of its own.

It also has to be measured honestly. Bing Copilot and Google publish per-query data; ChatGPT and Gemini do not. Surfaces that cannot be measured get marked “not audited” rather than implied — which is a discipline, not a limitation.

A rankings report cannot answer the only question that decides whether the programme continues. Search Console and GA4 are joined to the CRM so every cluster reports organic-sourced trials, paid conversions and MRR — and so the ones producing nothing can be cut without an argument.

The second half of this stage is classifying pages by the job they do, because a single site-wide click-through rate hides more than it shows. A fact page and a recommendation page succeed in different ways and have to be judged against different targets.

This is where the method closes: the clusters that produced revenue get more investment, the ones that produced sessions get cut, and the next quarter is planned from the result rather than from a fresh keyword export.