AI Visibility: What the Numbers Mean

AI visibility is how often AI assistants such as ChatGPT name your business when someone asks the kind of question your buyers ask. Revnu asks those questions on your behalf, on a schedule, and reads every answer. The results are on the Visibility pages of the dashboard. This is the measurement side of what's called AEO (answer engine optimization) or GEO (generative engine optimization). What are AEO and GEO? explains the idea.

In plain words:

• Visibility is how often AI answers to your buyers' questions mention you.

• Share of Voice is how much of the talk about your category is about you, not your competitors.

• Citation Share is how often the links in those answers point at your own pages.

The numbers, read correctly

Every number below counts measured answers: one AI engine's answer to one of your tracked questions on one run.

Three things trip people up:

• Visibility and Mention Rate can disagree. Visibility weights every topic equally. Mention Rate is the raw share for one question. A strong question in a big topic moves Mention Rate more than Visibility.

• A good position can hide a low visibility. Position only counts answers that mention you. You can average 1.5 and still be missing from most answers.

• Top-1 Rate and Avg. Position tell different stories. Say 6 of 10 answers mention you and 3 of those name you first. Visibility is 60%, Top-1 Rate is 30%, and Avg. Position might be 1.5. Top-1 Rate counts every answer; position counts only the ones that name you.

• Citation Share can be low while you're mentioned. An AI engine can name you while linking to a review site or a list. Citations to old domains, aliases or a separate blog host only count if you add them under Your Domains on the AI Citations page.

The headline numbers always use ChatGPT, Claude, Gemini and Perplexity, so they mean the same thing on every plan. That set is called Core 4 in the Platform picker at the top right of the Visibility pages. Pick one engine to see it alone, or All Platforms for every engine your plan measures.

Enlarge

A worked example: which number to report

Say Overview shows 4% visibility with Last 7 Days selected, but when you hover one day on the Daily Visibility chart it shows 8%.

• 4% is the whole window: of every answer measured in the last 7 days, 4 in 100 mentioned you.

• 8% is one day. On Pro most measurements happen on your weekly day, so single days jump around.

Report the window number, with the window: "4% visibility over the last 7 days, Core 4." Compare it with the same window, the same engines and the same questions next time.

Is this the same as share of voice in SEO tools?

They measure different things. SEO tools measure share of voice from Google search rankings and estimated clicks. Revnu's Share of Voice counts mentions inside AI answers: your mentions against your competitors' mentions. The two can move in opposite directions, so don't compare one with the other.

Revnu's Google numbers are on Visibility > Google Search, from your own Search Console. Your links and link authority are on Visibility > Backlinks.

Which engines Revnu checks, and how often

On Pro, the AI Visibility page shows "Checked weekly" and the weekday under each engine. That's when the next measurement runs. Your first numbers show up within a week on Pro, and within a few days on Managed.

Grok, Google AI Overviews, Google AI Mode and Copilot are measured on Managed only. On Pro they aren't measured at all, and the Pro numbers don't estimate them. Pro also has no results by region or persona. To get those and the extra engines, see Pro and Managed.

The Daily and Weekly toggle on the AI Visibility page only changes how the chart groups the data. It doesn't change how often Revnu measures.

A question you've just added is measured straight away, usually within minutes, on the engines your plan covers. If you add a lot at once, the rest are picked up in the next day's run. After that, on Pro, it's measured with the others on your weekly day.

You can't re-run a measurement on demand, for example right after you ship a fix. When you ask Revnu about visibility, it reads the stored answers and tells you their dates. To judge a fix fairly, leave your tracked questions as they are and compare the next scheduled measurement with the one before.

What each Visibility page shows

Search expansions and citations are where the next actions come from. How AI engines choose what to cite explains both, including the five Citation Categories.

Branded and non-branded questions

Revnu tracks non-branded questions: the ones a buyer asks before they know you, like "best invoicing tool for freelancers". That's where being named wins customers. If you try to add a question that names your brand, Tracked Prompts refuses it.

Revnu still asks branded questions in the background, but only to read how AI engines describe you. That shows on the Sentiment page.

What if my name is also a common word or another company's name?

Revnu checks each answer against your name, your domains and a description of what you do. An answer about a same-named agency, place or product doesn't count as a mention. Adding every domain you own under Your Domains helps it tell you apart.

If you still see mentions or citations that are about another product, tell Revnu which ones: "the Acme answers about the music app aren't us." It can reword or drop the questions that keep matching the other product. Write questions the way a buyer would, without your name, in a few words, for example "invoicing tool for freelancers".

Add, edit and remove questions

1. 1

   Open Tracked Prompts

   Open Visibility > Tracked Prompts. Clear any topic filter first: the box is off while one is set.

2. 2

   Type a question and click Add

   Write it the way a buyer types into ChatGPT: short, four to nine words, with no brand name. For example "invoicing tool for freelancers in the UK". Long, descriptive sentences make poor questions because nobody asks them.

3. 3

   Edit or remove one

   To change a question, click the pencil on its row. To stop tracking it, click the X.

You can also ask in chat: "track 'best CRM for dental clinics'", "stop tracking the pricing question", or "add Acme as a competitor".

With Last 7 Days or Last 30 Days selected, a small tag next to each question shows where it came from:

• Manual: one you added.

• Research: one Revnu found while researching your market.

• GSC: a real search from your Search Console.

• Keyword: a keyword you asked Revnu to target.

Removing a question stops new measurements of it. Keep hard questions on the list rather than swapping them for easy ones: a score you raised by dropping hard questions tells you nothing. Changed what you sell? Swap the old product's questions for the new one's, as changed what you sell explains.

Turning gaps into actions

The Improve Visibility card on Overview lists the gaps worth looking at, with a count of how many are To Investigate. Each one has an Investigate in Chat button. Click it and Revnu opens a chat about that gap: it re-reads the latest answers, picks one buyer question, and proposes one next step. Once a chat is open, the button reads Open Conversation.

Investigating only reads and advises. Revnu asks before it creates anything for you to approve, changes your tracking or drafts content.

The next step is usually one of these:

• Add what's missing to a page. If the answers miss facts that your pages don't state either, Revnu proposes the exact addition and the page it belongs on. Writing it follows your Blog Posts setting.

• Get listed where the answers look. If the answers cite a directory or a "best X" list that names your competitors, Revnu can submit you there. See free directory listings.

• Track a clearer question. If the wording leads AI engines to a different kind of product, Revnu proposes a better question to track next to the old one.

Each action says what the next measurement should show, so you can tell whether it worked. Changes take weeks to show in AI answers, not days.

Investigate: Track more queries

Track more queries shows up on the card when you track fewer than 25 questions, with a line like "You're watching 12." It means your score rests on a small set of questions, so one answer swings it a lot. It isn't an error.

Click Investigate in Chat and Revnu reads the questions you track and your business, then proposes up to three new buyer questions. Each one sits next to the nearest question you already track and says what new buying decision it covers. Nothing is added until you say so. Reply "track those" (or name the ones you want) and Revnu adds them. You can also add them yourself in Tracked Prompts.

How to pick the right prompts goes deeper. In short, approve the ones a real buyer would type before they've heard of you, the way they'd type them: four to nine words, no brand name. "What do I have to ask for us to be mentioned?" has no single answer. Pick the questions your best customers asked before they found you. A new question is usually measured within minutes, then with the others on your weekly day on Pro.

What Revnu does and what you do

Revnu doesn't email list authors or editors, and it doesn't buy or trade links. Where a page is only reachable by asking a person, it gives you the link and the pitch, and you decide whether to send it.

Track the competitors you lose deals to

The Competitors list starts from the brands AI answers name next to you, not from your website or your sales notes. A rival nobody's AI answer mentions won't appear until you add it, and Share of Voice only counts the competitors you track.

To add one:

1. Open Visibility > Competitors.

2. Type the name in the Add competitor row at the bottom of the table and click the plus.

3. To stop tracking one, hover its row and click the X (Hide Competitor).

Or ask in chat: "add Acme as a competitor." You can track up to 30. Your own brand can't be added as a competitor.

After you add one, Revnu re-checks your tracked questions straight away, usually within minutes, so the new competitor shows up in the table. After that it's measured with everything else, weekly on Pro.

Hand AI-visibility fixes to your coding agent

If a coding agent works on your site, it can get the fix list from Revnu and make the changes. Two ways:

• Give the agent a token. In Settings > Developers, mint a token with agent:message and results:read, labelled with the agent's name. The agent then asks Revnu through the ask_revnu tool or the API, for example "what changes do you need on our site to show up in AI answers for 'invoicing tool for freelancers'?", and makes them. Setup: Developers.

• Connect GitHub. Revnu opens the pull requests itself, and your coding agent reviews and merges them like a person's. See Connect GitHub.

Read the numbers alongside your pipeline

Visibility tells you how often you're named when buyers ask. Pair it with these to see the full picture:

• Demand. Visibility measures whether you're named on the questions you track. Search Console shows how many people search for those topics on Google.

• Visits. With the Revnu Tag on your site, ask Revnu "how much traffic did AI engines send us?" It reports visits by engine and landing page. Some engines hide where a visitor came from, so treat the number as a floor.

• Cause. Revnu compares the same question on the same engine before and after each change, and tells you when a result is too small to call. Engines also change on their own, so judge over several measurements.

Citations are up but leads are flat

Being named more often doesn't turn into inbound straight away. Buyers who meet you in an AI answer often come back days later, by searching your name or typing your address. To see the link:

1. Install the Revnu Tag and set up outcomes for trials, purchases and booked calls, so Revnu can count what visitors do.

2. Ask Revnu "how much traffic did AI engines send us?" It reports visits by engine and landing page.

3. Ask new leads where they heard about you. Add a "How did you hear about us?" field to your sign-up or booking form, with ChatGPT and other AI assistants as options. It's the only way to catch people who read an answer and came back later.

4. Give it time. Inbound from AI answers builds slowly. Judge it over months, not weeks, against the same questions and the same windows.