The GR8ER2 AI Visibility Standard

The Gr8 Score is a published, reproducible measure of how often AI assistants recommend a brand — and, unlike the visibility scores we have found sold across this category, it ships with the questions it was built from, the number of times it was run, and its margin of error.

It exists because a single AI visibility check is not a measurement. The sources an AI engine cites change substantially from one day to the next — 74% of the domains ChatGPT cited changed between consecutive days, and 61% on Google's AI Overviews. A number produced by asking once carries an error bar wide enough to swallow the finding. This Standard is the discipline of refusing to sell a single reading as if it were the whole picture.

Source: Detailed, Glen Allsopp, "I Tracked 70K+ ChatGPT and AI Overview Responses…", 23 Sep 2026 · 70,000+ responses over 28 days, non-logged-in US. Author discloses working at Ahrefs. · checked 2026-09-29

What the Gr8 Score measures

The Gr8 Score is a single number from 0 to 100, built from three measured components:

Mention Rate — how often the brand is named at all. Share of Voice — of the brands named alongside it, how much of the attention is the brand's, weighted by where in the answer it appears, because a first mention counts for more than a ninth. Citation Rate — how often the brand's own pages are cited as a linked source, not just mentioned in passing.

The three are combined with a published, fixed weighting — 40% Mention Rate, 35% Share of Voice, 25% Citation Rate — and the components are always shown separately, so anyone can re-weight them for their own context. There is no secret formula. The weighting is a stated judgement, frozen per version, not a number we ask you to take on trust.

The Prompt Corpus (the owned term)

A Prompt Corpus is the specific, published set of buyer questions a visibility score is calculated from. A score without its Corpus is not a finding — it cannot be compared with any other score, and it cannot be checked.

This matters more than it sounds. Applying two defensible reweighting schemes to one published dataset produced results of 39.7% and 70.5% — a swing of more than thirty points — without changing a single underlying rate. Whoever writes the questions decides what “winning” means. So the Standard requires the Corpus to be published in full with every Score, frozen for the duration of a measurement cycle, and versioned rather than quietly edited.

Source: arXiv:2609.06811 · revised 6 Sep 2026 · single-author reanalysis, not peer-reviewed (author attribution recorded but not independently re-verified, ROI Lab run 1). · checked 2026-09-29

How it's measured

Because a single reading is a sample, every Gr8 Score is run repeatedly and reported with its confidence. The floor is not ours to invent: the independent St. Gallen study — titled, pointedly, Don't Measure Once — explicitly recommends at least seven runs per question for brand visibility, and at least eight when source-level coverage matters. Because the Gr8 Score captures citation, eight binds. The result is reported as a distribution — “named in eight runs out of eight”, not simply “named”.

Source: Schulte, Bleeker & Kaufmann · arXiv:2604.07585 · Apr 2026. · checked 2026-09-29

Two rules keep a reading honest over time. A before-and-after figure is only reportable when both readings used a byte-identical protocol, with both run counts printed beside the change — raising the run count between two measurements manufactures a movement that was never there. And not every answer reaches the live web: in one survey, 57.8% of ChatGPT repetitions did not trigger a web search at all, so the Standard reports the no-retrieval rate as a result in its own right rather than quietly discarding it.

Source: Critical survey of 45 GEO studies, arXiv:2607.14035, 15 Jul 2026 · single-author survey (author attribution recorded but not independently re-verified, ROI Lab run 1). · checked 2026-10-01

Engines are measured and labelled separately, never silently blended — they differ enormously, from around 15 sources cited per answer on ChatGPT to around 3 on Gemini. And no number is ever invented: every figure traces to the measurement that produced it or a sourced, dated record. A brand that scores zero is shown as zero, not dropped.

Source: Semrush 2026 AI Visibility Index · 126M US prompts · 26 Jun 2026. US-only. · checked 2026-09-29

What we publish that no one else does

Across every AI-visibility vendor and agency we have surveyed, three things go unpublished by all of them. The Standard requires GR8ER2 to publish each:

Our own score. GR8ER2 measures its own Gr8 Score every cycle, to the same standard as any client — including when it is bad. The error bar. Every Gr8 Score ships with its confidence band. A number without one is a snapshot wearing a score's clothes. The run count. We state how many times each question was asked. Most of the market runs it once and never says so.

Our current reading is an open-field score of zero.See our own number →

What the Gr8 Score is — and isn't

The Standard measures visibility honestly; it does not claim to know the single lever that changes it. A critical survey of 45 GEO studies found that no reviewed technique shows a stable, cross-platform causal effect on discoverability. Anyone selling a guaranteed method is ahead of the evidence.

Source: arXiv:2607.14035 · 15 Jul 2026 · single-author survey (author attribution recorded but not independently re-verified, ROI Lab run 1). · checked 2026-09-29

ROI sits even further from the evidence than visibility does. The same survey, read at its primary source, is blunt: claims about GEO return on investment clearly outstrip the academic evidence. The one quasi-experimental traffic estimate it names — a 1.82× multiplier — it calls suggestive rather than causally established, because a conservative placebo test fails. That is the honest floor under every ROI promise in this category, and it is why the Standard measures what AI says about a brand and treats the return as a separate, carefully-bounded question.

Source: Critical survey of 45 GEO studies, arXiv:2607.14035, 15 Jul 2026 · single-author survey (author attribution recorded but not independently re-verified, ROI Lab run 1). · checked 2026-10-01

What the evidence does support is where visibility correlates: across 75,000 brands, off-site branded mentions predicted AI visibility far better than owned-site authority — 0.664 against 0.266. So the Standard measures the outcome and treats the work to move it as a separate, evidence-limited craft — never a promise baked into the number. A Gr8 Score is a dated observation, read as a line over time, not a permanent rank.

Source: Ahrefs · 75,000 brands · Spearman · 12 Dec 2025. Correlation, not causation. · checked 2026-09-29