Methodology

How we actually measure this

AI visibility is how often, and how prominently, your brand shows up when someone asks an AI engine a question your buyers actually ask. A citation is when that engine names your brand, links your domain, or draws directly on your content to answer — not just a passing mention of your name in the response text.

ChatGPT Perplexity Gemini Google AI Overviews

// tracked continuously, not a one-time snapshot

Which engines, and why

We track where your buyers are actually asking

We track ChatGPT, Perplexity, Gemini, and Google AI Overviews because together they cover the large majority of AI-mediated search and chat traffic a brand realistically encounters today. Each engine retrieves and ranks sources differently — Perplexity leans heavily on live web retrieval, Google AI Overviews draws on the same index that powers organic search, ChatGPT blends training data with browsing when it's enabled, and Gemini pulls from Google's index plus its own grounding layer.

A brand that's cited reliably in one engine and invisible in another isn't done — it's half-measured. So we check all four for every client, every cycle, not just the one that's easiest to test.

Definitions

Citation vs. mention vs. recommendation position

Strongest signal

Citation

The engine credits your brand as the source — a named link, an attributed passage, or your domain listed as a reference. This is the signal we weight most heavily in reporting.

Worth tracking

Mention

Your brand name appears somewhere in the answer, but not as the credited source — for example, listed in passing alongside competitors without a link or clear attribution. Not the same as being cited.

Where you land

Recommendation position

Where you land when an engine names multiple options — first-named, mid-list, last-named, or omitted. Most users act on the first one or two options an engine surfaces.

Cadence

Monitoring is continuous. Active work is deliberate.

These are two different clocks, and conflating them is where most vendors get vague. Monitoring runs against as many prompts as are relevant to your business — we don't cap it at a round number, because a thin prompt set gives you a false picture of where you actually stand. If your buyers ask AI engines twenty different versions of "best X for Y," we track all twenty, not three.

Active work — the content, technical fixes, and structural changes meant to actually move those results — runs on a tighter, deliberate cadence: 2-3 pieces of work per week, roughly 8-12 per month. That pace isn't arbitrary. It's set by how fast a change can realistically be written, published, re-crawled, re-indexed by the relevant engine, and re-tested against the prompt set without us guessing whether it worked. Shipping faster than that just means publishing before anyone can tell what's working.

This is the full explanation — Strategy and Full Execution both link back here instead of re-explaining it.
Proof, not promises

Sample citation log

This is a realistic sample, not a live client account — it's here to show the format, not to claim a specific result. Every client gets a log built the same way: real prompt, real engine, real outcome, logged and dated.

Prompt Engine Result
"best AI visibility monitoring tool for agencies" Perplexity Cited
"how to get cited by ChatGPT" ChatGPT Not cited
"GEO vs SEO explained" Google AI Overview Cited
"answer engine optimization services" Gemini Not cited
"what is a content brief for AI search" Perplexity Cited

Sample log for illustration. Client logs are dated, scoped to their own prompt set, and delivered monthly.

What we don't claim

No guaranteed rankings. No fabricated data.

No one can guarantee a specific citation, ranking, or recommendation from ChatGPT, Perplexity, Gemini, or Google — these are third-party systems we don't control, and their outputs change. Anyone promising a guaranteed placement isn't being straight with you. What we commit to instead: we report exactly what we measure, on the cadence we say we measure it, and we never fabricate a result to make a report look better. If a prompt set shows no movement in a given month, the report says so.

It's also worth naming plainly: some services marketed today as "GEO retainers" are, underneath, standard SEO work with new terminology on the invoice — keyword-driven content and backlink building, rebranded for a buyer who's heard "AI search" and wants to act on it. That work isn't worthless, but it's not the same discipline as tracking and improving how AI engines cite and recommend a brand specifically. We built our monitoring and reporting around the engines themselves — the citation log above is what that looks like in practice, not a renamed keyword report.

One more thing we don't do, on either program: link building or backlink generation, in any form. Our work stays on-page and on-site — no outreach, no guest posts, no paid placements.
Go deeper

Related reading

Process

How it works

The 12-agent system and the 5 human approval gates that sit between any output and your site.

See the pipeline →
Reference

Glossary

AEO, GEO, and the rest — plain definitions, no jargon left unexplained.

Read the glossary →
Cost

Pricing

Strategy and Full Execution, side by side, with the cadence above already priced in.

See pricing →
FAQ

Common questions

What counts as a citation vs. a mention?

A citation is when an AI engine names your brand, links to your domain, or pulls a passage from your content as part of its answer to a real prompt. A mention is looser — your brand name appears in the response text without being credited as the source.

How often do you check AI visibility?

Monitoring runs continuously across as many prompts as are relevant to your business. Active work — the changes meant to move those results — runs 2-3 pieces per week, roughly 8-12 per month.

Do you guarantee rankings or citations?

No. No one can guarantee a specific citation, ranking, or recommendation from a third-party AI engine. We report what we measure and never fabricate a result — and we don't do link building or backlinks on either program.

Want to see how this would run for your prompt set?

Book a call