Generative Search Signals

Generative Search Signals is our name for the signals that decide whether an answer engine can retrieve a company, understand what it is, and find something on its site worth quoting. It is the mechanism inside our SEO program, not a separate product.

Why a framework exists for this

Because the failure modes are different from the ones traditional SEO diagnoses. A page can rank well and still be uncitable. It can be well written and still unreadable to the systems doing the citing. It can be accurate and still fail to establish who wrote it.

Ranking answers the question "where does this page sit in a list". Citation answers a different one: "is this passage the best available source for this specific claim". Those come apart, and a framework that only measures the first will miss the second entirely.

The three signal layers

Retrievability, entity clarity and citable evidence. They are ordered deliberately: each depends on the one before it. A brilliant, quotable passage on a page no crawler can read is worth nothing.

Layer one: retrievability

Can the systems that build answers actually get the page, and is the company present in the indexes each one reads from?

The most common failure here is JavaScript. Most AI crawlers do not execute it and they work on short timeouts. Content that only exists after a script runs is invisible to them while looking perfect to you. The diagnostic takes ten seconds: open View Source rather than Inspect, and look for your headline in the raw HTML.

The second most common failure is a robots.txt written as a whitelist. Blocking everything and then naming a few permitted crawlers silently excludes every AI agent nobody thought to list, including ones that did not exist when the file was written. Allowing by default and blocking deliberately is the safer posture.

The third is index coverage. Different engines read from different places. Being healthy in Google says nothing about Bing, which is the route into ChatGPT and the one most companies never check.

Layer two: entity clarity

Can an engine tell who this company is, distinguish it from similarly named ones, and find the same facts wherever it looks?

Entities are established by corroboration, not assertion. A company cannot declare itself into a knowledge graph. What it can do is make sure that every place it appears agrees: the same legal name, the same description, the same address and phone number, the same founding year.

Inconsistency is the quiet killer. A business whose address is formatted three different ways across its own site, its business profile and its directory listings has given engines a reason to doubt all three. Reconciliation fails and the entity stays fuzzy.

Layer three: citable evidence

Is there anything here worth quoting?

Engines slice pages into passages and score each one independently against a question. That changes what a good page looks like. A section that only makes sense after reading the two above it will not survive extraction. A claim buried in the sixth paragraph loses to the same claim stated cleanly in the first.

It also raises the bar on substance. Content that restates what already ranks adds nothing an engine needs. What earns a citation is material that exists nowhere else: a figure you measured, a specification only you publish, a result with a name and a number attached.

The work in practice

How it is measured

By asking the engines. We run the questions a company's buyers actually ask across multiple answer engines and record what comes back: whether the company is named, how prominently, in what context, and which sources the answer drew on instead.

That last part is the useful one. Knowing a competitor was cited is mildly interesting. Knowing which specific page was cited tells you exactly what to go and earn.

Measurement runs on a schedule rather than once, because the answers move. A single reading is an anecdote.

What this framework does not do

It does not guarantee citation. No framework does, and nobody controls what an answer engine says. Engines change their outputs continuously, and a company can do everything right and still not be named for a given question on a given day.

It also is not a replacement for the rest of search. Rankings still matter, still drive traffic, and are still an input to what gets cited. This works alongside the traditional program rather than instead of it.

And it is not fast in every layer. Retrievability fixes can land in days. Entity corrections often show within weeks. Building genuinely citable material is content work and takes as long as content work takes.

Why it is not sold separately

Because on its own it produces visibility without a pipeline. A company that is beautifully cited and has no organic presence, no working conversion path and no measurement has bought a metric rather than a result.

The signals above are also, layer for layer, the same work that earns rankings: clean rendering, clear entities, well structured and genuinely useful content. Selling them as a separate line item would mean charging twice for one program and inviting a client to buy half of something that only works whole.

Frequently asked questions