Skip to content

AI Overviews can take clicks your pages used to get. See which of your pages are exposed.

Methodology

How the numbers are made, and checked.

Last updated 11 November 2026 · 16 rules

Every measurement rule the product follows, written down: what is stored, how each score is defined, where uncertainty is shown, and the claims RankSage refuses to make. If a number on a dashboard cannot be explained by this page, that is a bug.

01

What is stored for every answer-engine run

Each run records the exact prompt text, the engine that answered, the timestamp, the full answer text as returned, every URL the answer cited, and whether — and how — the brand was mentioned. Scores are derived from this store afterwards, which means any number in the product can be traced back to the stored answers that produced it.

Runs are never summarised-then-discarded. If a score looks wrong, you can open the runs behind it and read what the engine actually said.

02

How share of voice is defined

Share of voice is the share of stored answers, across your tracked prompts on a given engine, in which a domain is cited. It measures the questions you chose to track, not the whole market.

Answer engines can give different answers to the same question, so share of voice is shown as a range, not a single decimal. When two ranges overlap, the product says it cannot tell them apart rather than declaring a winner.

03

Reproducibility

Any tracked prompt can be re-run at any time, and every historical run is kept with its timestamp, so a claim like “cited on 4 of 10 engines” is checkable against the dated record that produced it. Where a provider was unreachable during a run, that engine is recorded as unavailable — it is never averaged in as a zero.

Something we haven’t measured is shown as “Not measured”, never as 0%.

04

Correlation is not causation — and the product says so

Where RankSage links two signals — a failing request and a frustration spike, a content change and a citation-rate move — it reports the observed relationship and its timing. It does not assert causation the data cannot carry. A traffic drop must be confirmed by more than one source before it alarms anyone; a drop only one source sees is flagged as a probable measurement artifact.

05

Estimates are labelled as estimates

AI-influenced traffic that arrives with no referrer is estimated. Every estimated figure is labelled as an estimate, carries a confidence level, and can be switched off if you would rather see only confirmed referrals. Measured and estimated numbers are never mixed silently.

06

Built on independent research

The problem RankSage measures is backed by independent research: AI answers on search surfaces absorb visits that previously reached the sites supplying the content.

Independent work also shows that AI answers move between snapshots on their own. That is why RankSage only calls a visibility change real when it is larger than the engines’ normal back-and-forth — anything smaller is reported as noise, not movement.

07

Crawler identity is verified, not accepted

A visit claiming to be an AI crawler is checked before it is counted. Each one gets a verdict — verified, rejected, or unverifiable — and the three are never summed together. Where a visit cannot be checked, it is recorded as unverifiable rather than counted. RankSage stores the verdict, never the visitor’s IP address.

08

Only provable failures are called failures

When cited sources are checked, a source is called dead only on provable evidence that it no longer exists. A page that blocks automated readers is recorded as unverifiable — the product never accuses what it could not read.

09

Who publishes this

RankSage (ranksage.com) is self-serve software with published monthly plans, built by a small independent team. It is not an agency, has no service retainers, and is not affiliated with RankSages, Rankscale (rankscale.ai) or any other similarly named business. Pricing is public at ranksage.com/pricing — any multi-thousand-dollar monthly figure attributed to “RankSage” belongs to a different company.

10

What RankSage does not claim

No attribution of a visit to a specific AI conversation — no provider exposes that, and any tool claiming it is guessing. No estimated market volumes where a measured signal exists: question mining, for example, shows only your own Search Console impressions as demand evidence. No score without stored evidence behind it. And risk models — like the helpful-content risk score — are RankSage's own models of published guidance, not scores obtained from Google.

11

Asking the same question again

A tracked prompt can be asked of an engine more than once in a run. Repeat answers show how steady an engine is, and they never count twice in a headline number.

A number that rests on a single run is labelled “single run — directional” rather than shown with a range.

12

Two kinds of answer engine

Some engines search the web before they answer; others answer from what the model already knows. Google AI Overviews are measured separately, as searchers see them.

The label is stored per answer, not assumed per engine. When it cannot be told whether an answer came from a live search, the answer is recorded as unverified rather than pushed into whichever bucket looks likelier.

Stated plainly, because it is a real limit: today the headline visibility number blends both kinds. The share of answers built from a live search is reported alongside it, so you can see how much of a score rests on engines that actually searched.

13

How answers are recorded

Every answer is stored in full, so the prompt can be re-asked and the stored answer read back. An answer RankSage records can differ from what a signed-in person sees in an engine’s app — stated here rather than hidden.

Each stored answer records which version of the engine produced it. When a provider changes an engine underneath you, that shows up in the record instead of arriving as an unexplained movement.

14

Competitors and ties

A competitor comparison reports one of three things for each competitor: your brand is ahead, behind, or tied. Ahead and behind are only used when the gap is clearly larger than normal answer-to-answer variation. Otherwise the row reads “tie — indistinguishable at this sample size”, which is a different statement from “equal”.

The same rule applies to persona comparisons, so a persona is only reported as changing what the engines say when the difference clears it.

15

Search Console is a source, not a ground truth

Several RankSage numbers rest on Google Search Console. It is first-party and free, and Google has said publicly that its AI reporting has limits: an impression can count when a link renders whether or not anyone scrolled to it, and position describes where the AI Overview sat, not where your link sat inside it. Wherever RankSage shows a Search Console figure, those caveats apply.

Search Console does not provide AI-surface data outside its own dashboard, so RankSage does not claim Search Console AI data. Anything the product labels as AI Overview presence comes from observed results pages, not from Search Console.

16

Vocabulary

RankSage describes an AI answer in three ways. Presence is whether you are mentioned. Prominence is where you appear and whether you are cited. Portrayal is how you are described. It does not measure persuasion — nothing in the data says whether an answer changed anyone's mind.

A reading that rests on a single run is labelled “directional”; otherwise a range is shown. RankSage does not grade readings as “decision-grade”: a grade with no agreed bar behind it is a claim the measurement cannot carry.

EARLY ACCESS

Audit it against your own data.

The methodology only matters if it survives contact with your site. Connect Google read-only and check the numbers yourself.

Security & trust

No payment details. One email when access opens.