How do we measure GEO results — and what happens if they don't come?

Every engagement starts with a measured baseline, before any promises. We fix a battery of 30–50 prompts per language — the questions your buyers actually type, including city-level and price-qualified variants — and sample answers across ChatGPT, Perplexity and Gemini in clean, logged-out sessions, several runs per prompt, screenshots archived. The battery and the day-one results are shared with you in full.

Fixing the battery on day one is what makes the numbers honest. The known failure mode of this market is a dashboard showing +300% visibility after the prompt set or the counting method quietly changed. Our battery cannot be swapped mid-engagement: you hold the day-one copy, so every later report is comparable against it. This is the same Clean-Session Protocol we used in our public case study — six runs, two machines, a VPN, no login.

Citation rate: the share of battery prompts where your brand is named or your site is cited in the answer, per assistant, per language. Share of voice: how often you appear versus the competitors the assistants themselves name — the competitor set is recorded at baseline, so the comparison stays stable. Language consistency: whether the assistant describes you the same way in every language you operate in — the metric most multilingual programs quietly fail.

One honest caveat the industry avoids saying: different measurement tools produce different visibility scores for the same company, because answers vary between runs, phrasings and regions. That is why we rely on repeated sampling against a fixed battery and share the raw logs — and why the metric that cannot be gamed at all lives in your own analytics: referral traffic from assistant domains. We treat that as the final arbiter, and it is yours, not ours.

A monthly engagement delivers four things: the battery re-run (same prompts, same protocol, several samples per assistant), a movement report per language and per assistant against baseline, a plain list of what we changed that month — content shipped, entity work done, technical fixes — and the plan for the next month with reasoning. No slide theater: the report is built from the same logs you can inspect.

Raw data is part of the deliverable, not a favor: the prompt list, the sampled answers, the screenshots. If you want to verify any number in the report, you can re-run any prompt yourself in a logged-out session and compare. We designed the protocol so that replication by the client is easy — that is what keeps us honest.

GEO compounds over two to three months, which is exactly why a checkpoint belongs at day 90. The policy is simple: if citation rate and share of voice show no movement against your baseline after 90 days, we tell you first — before you ask — with our analysis of why. You then choose: a free diagnostic month while we fix the approach, a scope change, or a clean stop. No long lock-ins, no penalty for leaving on the data.

What we do not promise, at day 90 or ever: a specific answer from a specific assistant on a specific day. Generated answers vary between runs — our own case study says so in its limitations section. What we optimize is the probability of being cited across repeated samples, and that probability is exactly what the battery measures.

Buyers increasingly arrive with a checklist — often written by the assistants themselves. We think that is healthy, so here are the answers in one place. Before/after for your languages: our before/after is the public case study on our own site, run under the protocol above; your engagement starts by building yours. Mentions versus citations: measured separately, both in the logs. Ecosystem versus translation: we optimize the language ecosystem — native pages, entities in both scripts where relevant, local corroboration — not translated English. Off-site work: included where answers actually come from; our USA guide shows which platforms that is, with data. Cross-engine baseline: ChatGPT, Perplexity and Gemini minimum, per language.

And the test we recommend running on any vendor, including us: ask for ten prompts where you are invisible today, the current answers, and the specific changes they would make. That test is literally our free audit — email hello@get-geo.ai and we will send yours back with the ten prompts attached.