AI Assistants & GEO report

Who this is for: marketers working on how their website shows up in AI assistants and answer engines (GEO and AEO), and digital agencies who report that work to their clients.

What you'll learn: what the AI Assistants & GEO report shows, why it reads differently from your analytics and from tools that track AI mentions, how to run it in your AI assistant, how to share it with a manager or a client, and how to take it further with your own questions.

Most GEO and AEO work is measured at the top: was the brand mentioned, was the page cited, how often. That tells you whether an assistant talked about you. It doesn't tell you who clicked through, whether they were Focused visitors, or whether they became Engaged Leads. The AI Assistants & GEO report starts from the visit: which assistants send you people, which pages they cite, how many of those visitors behave like buyers, and which AI crawlers are reading your website to decide what to cite next.

In short: The AI Assistants & GEO report is a ready-made report your AI assistant builds from your Bread & Butter data, for any period you choose. It shows the visitors that ChatGPT, Perplexity, Gemini, Microsoft Copilot, Claude and other assistants send you, how many are Focused visitors and Engaged Leads, which assistants and cited pages do the work, where those visitors are, which AI crawlers read your website, whether you have an llms.txt file, and how AI's share of your Focused visitors has grown. Every card explains what the chart says, why it matters and what to keep an eye on, and ends with a question to ask next. Run it monthly, send it to your manager or your client, and ask your assistant to dig into anything it shows. Owners, Administrators and Marketers can run it.

What the report answers

Do AI assistants send buyers, or only browsers? Every assistant and every cited page is measured by Focused visitors and Engaged Leads, and compared with your website as a whole, not only by visits.

Which pages are the assistants citing? The page a visitor from an assistant lands on is the page the assistant cited. The report lists them, shows which ones turn visitors into Engaged Leads, and which ones lose them.

What are the assistants reading? AI crawlers visit your website to learn what it says. The report shows which crawlers came, how many pages they read and when they last came.

Is the work paying off? The month-by-month and impact cards show whether AI's share of your Focused visitors and Engaged Leads is growing, which is the question your manager or your client will ask.

What's in it

The report opens with the period at a glance: AI visitors and their share of all visitors, the Focused Rate of AI visitors against your website as a whole, Engaged Leads from AI assistants, and the share of the assistant that sends you the most, each against the period before, with a short paragraph on what changed.

Then come the cards, each one on its own:

AI funnel. How visitors sent by AI assistants became Focused visitors and Engaged Leads, and where they drop off along the way.

By assistant. Each AI assistant in the period, with its visitors, Focused visitors and Engaged Leads side by side, its Focused Rate, and a Scale, Watch or Investigate verdict. Assistants are named from the referrer, the information a browser passes on when someone clicks a link in an answer.

Month by month. AI visitors every month since the first one, with Focused visitors and Engaged Leads on the same chart, so you can see whether the growth comes from more of your pages being cited or from more people asking.

Pages AI assistants send people to. The pages AI visitors land on, which are the pages the assistants cite, each with its visitors, Focused visitors, Engaged Leads, Focused Rate and a verdict. Under the chart, Cited pages that lose the visitor lists the pages assistants still send people to that no longer answer the question, such as an out-of-date post or last year's event page, with what to do about each one.

Where AI visitors are. AI visitors by region, with the Focused Rate of each. A region that shows up more among AI visitors than across your website is a question your content answers well.

AI crawlers reading your website. Visits from AI companies' crawlers in the period, such as GPTBot, ClaudeBot and PerplexityBot: how many pages each one read, which pages, and when it last came. Crawlers are recognised by the user agent they declare. They are bots, so they are kept out of every visitor number in this report and in Bread & Butter.

llms.txt and what to do next. The llms.txt check for your website. llms.txt is a proposed convention: a short text file at the root of a website that tells AI assistants what the site is and which pages answer which questions. The card shows "Not found", or the score and date of the last scan. Below it, up to three suggestions from your AI assistant on what to do next, drawn from the cited pages, the crawlers and the funnel in the report, each with what it is expected to move. Two tiles track the results: how many pages assistants cite (pages with 50 or more AI visitors) and the one figure the suggestions are meant to change. A box on the card says, word for word: "These suggestions come from your AI assistant, not from Bread & Butter. Bread & Butter reports the data and provides it to your assistant; it does not recommend or change anything on your website. Different AI assistants can reach different suggestions from the same data, and the same assistant can answer differently on another day. Nothing changes until you change it."

Impact to date. Focused visitors from AI assistants each month, and their share of all Focused visitors on your website, with the first and latest month side by side and the Engaged Leads from AI as a share of all your Engaged Leads.

Under every card are three short sections: What the chart says, the three findings that matter; Why it matters, in a sentence or two; and What to keep an eye on, the figures to watch next period and the level at which each one needs a look.

Why it reads differently from your analytics and your AI visibility tools

It counts people, not hits. Every number leaves out bots. Focused visitors are the real people reading your website: everyone who has become a lead, plus every visitor scoring 6 or more on Focused Score. See Why does the Focused Tab show so much less traffic than my analytics?

It starts where AI visibility tools stop. Tools that track mentions and citations tell you whether an assistant talked about you. This report tells you what happened after: who clicked through, whether they were Focused visitors, and whether they became Engaged Leads. For the difference between AEO and GEO, and why the visit after the citation matters, see AEO vs GEO: what they are, how they differ, and what happens after an AI sends you a visitor.

It names only what it sees. Assistants are named from the referrer, never guessed, and the report never invents the question someone asked, because assistants don't pass it on. Some AI visits arrive without a referrer, for example when someone copies a link out of an answer, and count as Direct / Unspecified, as they do in every analytics tool. The real AI number is a little higher than the one shown.

It shows who is reading, as well as who is visiting. Your analytics leaves crawlers out and stops there. This report keeps them out of every visitor number and shows them on their own card, because a crawler reading a page is the earliest sign of what an assistant will cite.

It tells you where to look. Scale, Watch and Investigate rank your assistants and cited pages against each other: the top third Scale, the middle Watch, the bottom Investigate. AI traffic can't be bought, only earned, so the verdicts point at content: the pages to repeat, and the pages to rewrite or redirect. Investigate means look closer, not switch it off: the decision stays with you, because you know things about your business the data can't see.

How to run it

Connect your AI assistant to Bread & Butter first. See Bread & Butter MCP server connections.

Finding the reports. Once you're connected, the reports are ready to run straight away; there's nothing to download or set up. Ask your assistant "What Bread & Butter reports can I run?" and it lists the reports your role can run, with what each one is for. The Insights tab in Bread & Butter lists them too, each with a prompt you can copy.

More than one account. If your login gives you access to several Bread & Butter accounts, name the one you want. If you don't, your assistant asks which one before it runs anything.

Coming back later. The connection stays in place, so there's no code or key to remember. Open a new conversation, check that the Bread & Butter connector is switched on for it (in Claude, under the + button, then Connectors), and ask for the report by name. If the connection has lapsed, your assistant asks you to reconnect first.

Ask for the report by name, with a period:

"Run my Bread & Butter AI Assistants & GEO report for last month."
"Run my Bread & Butter AI Assistants & GEO report for the last six months, for my main website."
"Run my Bread & Butter AI Assistants & GEO report for Q3 and compare it with Q2."

Your assistant gets the report from Bread & Butter, fills it with your account's data and shows it as a page. Each run uses credits like any other question. See How do Credits work across storage, syncing, and profile enrichment?

A good rhythm is monthly. Visits from AI assistants are still small for most websites, and a month gives enough to read. For a first look, run it for the last six months so the month-by-month card has a shape to show. To have it arrive on its own every month, see Run your Bread & Butter reports on a schedule.

For a marketing manager. The first card is the summary: four figures and a paragraph on the period. The impact card answers the question a manager asks about GEO: is it worth the work. The verdicts and the suggestions on the llms.txt card give the meeting its agenda.

For an agency reporting to a client. Run the report for each client's website on the first of the month and send it as it is. It explains itself: every chart carries its own reading, so the client doesn't need you in the room to understand it. The cited pages and the suggestions give your GEO retainer a monthly list of work, and the next report shows what moved. Each card can be printed on its own, for a slide or a monthly pack, and printed copies leave out the prompts and the schedule box. If your client's team works in Bread & Butter with the Sales role, they see their leads but not this report, which lets you go through the results with them first. See What are the different team roles, and what do they have access to?

Someone forwarded it to you? The box under the first card explains how to connect your own assistant and run it yourself.

Taking it further with your own questions

The report is a starting point. Every card ends with Try this next in your AI assistant, a question chosen from that card's most important finding, with a button to copy it. Ask it, or ask your own. A few examples:

Narrow it down. "Show the AI Assistants & GEO report for Perplexity only." "Run it for visitors from one province or state." "Show only the AI visitors who landed on the pricing page."

Go one level deeper. "Show Perplexity visitors by landing page with the Focused Rate of each, so I can see which pages it cites and repeat the format." "Show the pages each AI crawler read most in the last 30 days, with the AI visitors each page then received."

Compare. "Compare this month with the same month last year." "Show Focused visitors from AI assistants against Organic Search month by month, as a share of all Focused visitors, so I can see whether AI is taking from search or adding to it."

Find the pages to fix. "List every page with 20 or more AI visitors, with Focused Rate, Engaged Leads and the last time it was updated, so I can see which cited pages are out of date."

Act on it. "Write an llms.txt for my website from the pages in this report: what the site is, the main sections, and the pages not to cite."

Change the report. "Add a card with AI visitors by device." "Leave out the regions card." "Write the summary paragraph for a client who doesn't know what a Focused visitor is." "Translate the report into French."

Make it yours. Once a version works for you, ask your assistant to remember it, or save the prompt as a scheduled task so it arrives the same way every month.

Your assistant can only answer from the data your account has. When a card has nothing to show for the period, it says so instead of guessing, and a new account with no data yet gets a short note at the top with what to set up first.

What it needs

Bread & Butter installed on your website. Data starts with the next visit, from people and from crawlers. See How to Install Bread & Butter on Your Website

Visits from AI assistants. The report fills in when assistants send people to your website or read it. Focused Score works from the first visit; a monthly report reads best with a full month behind it.

User journey events, for the funnel. Name the actions that matter on your website, such as a quiz, a booking or a pricing page view, and the funnel and the suggestions can show how far AI visitors get. See What are User Events, and how do I use them?

Nothing for the crawlers or llms.txt. Bread & Butter's bot filter records AI crawlers on its own, and the llms.txt check runs on your website's address. After you upload an llms.txt file, ask for a rescan and the card shows the score and the date.