TL;DR
- Prompt Intelligence is live in Visibility on managed plans. Open any tracked prompt and I read every observed answer to it, pull out the claims, group them into recurring claims and answer patterns, and compare the AI systems against each other. Every finding links back to the answers it originated from.
- The Opportunities tab in Visibility is now an AEO work queue. Each row is a job: create a page, improve a page, get added to a page the AI systems already cite, correct a claim, or investigate a change. Each one includes evidence, a priority, a confidence rating, and a button that will start the work.
- AEO audits are faster - ohh ya! The small decisions inside an audit now run on the same fast decision system one model that picks tools in chat (Jev is very cool), and the review of each recommendation runs in parallel on GPT-6 Luna.
- Comparison dates now stays intact through the whole Visibility screen, with a new dashed line for the earlier period on every trend chart and a summary written as a comparison.
Prompt Intelligence
AI Search Visibility has been collecting how AI systems answer your prompts, run after run. There are loads of cool things you can learn from with this data. Expand a prompt in the Prompts tab and there is a Prompt Intelligence card. Click Generate intelligence report and I (Moose) will read each observed answer, extract the claims it makes, group them into recurring claims and answer patterns, and compare what each AI system says against the others.
The report opens with a Summary that includes how often your brand appears in the answers, the one claim the AI systems consistently attach to you, an AI system that includes you less than the rest, a way of answering the prompt that rarely includes you, and whether the top recommendation is contested between brands. Under that is the Supporting analysis, with the distinct answer patterns, the recurring claims ranked by how many answers carry them and how many systems agree, the differences between models, and an Answer behavior line for how stable the answer has been over time.
Every number has a View evidence link. It lists the observed answers behind that number, the claims in each one, and the full response. We wanted every finding traceable to real answers.
AEO Opportunities, as a work queue
The Opportunities tab in Visibility is rebuilt as a list of jobs. Every row is written as the action, such as "Create a page that answers this prompt" or "Get added to this page", followed by what the answers show, the next step, and the outcome to expect. Five kinds of job come out of it: create coverage, improve a page, earn inclusion on a page the AI systems already cite, correct the description, and investigate a change.
Each row has a priority with its reasons, based on how many prompt and AI system pairs it affects and how consistently the gap held across runs. Confidence is Strong, Directional, or Needs investigation, and an estimated effort (to complete the task) is rated as well.
There is a button on each row where you can trigger the work. The "Create brief" button opens the brief intake form with the prompt as the topic. "Audit page" button runs an AEO audit that knows which opportunity it came from, so it writes an improvement suggestion based on that opportunity. "Draft a pitch" button writes an email draft for you to consider using, or it writes a reply to a thread, or a claim-your-listing note, depending on the site. Sites that cannot be pitched, are marked to monitor.
You can mark a row done, dismiss it, snooze it, or leave a note for the team. A done or dismissed row reopens on its own when new evidence shows up. On managed plans, Prompt Intelligence findings attach to the row as evidence. The opportunity engine runs on your machine, and matching your own pages to a prompt uses a local embedding model.
Quicker AEO audits
An AEO audit makes a lot of small decisions on the way to its recommendations: what kind of edit a recommendation calls for, where on the page it belongs, what role a target query plays, what kind of search results a page is up against. Those now go to the same fast system one model (Jev) that picks tools in chat. A recommendation that clearly passes will skip the full review, and the rest are reviewed one at a time, in parallel, on GPT-6 Luna by default. This is on for paid managed plans and BYOK customers. On our test audits the whole run finished about 40% faster.
Also in the 0.6.65 build
Comparison dates now connect every part of the Visibility screen. Set one and every trend chart gets a dashed line for the earlier period, the Overview cards show the change, and the AI summary is written as a comparison. Topics & Engines shows the change per topic and per engine as well. Also, a bug fix, the comparison label showed the wrong day on some ranges.
The model behind chat's quick decisions, like what kind of brief you meant and which tool to plan, moved to GPT-6 Luna on paid plans. A better model at the same cost on your allowance.
When I suggest prompts to track in chat, each prompt now has a Create brief button beside Add prompt. It opens the brief intake right in the conversation with the prompt filled in as the topic. (This was user-suggested, thank you for that feedback!)
Also, another bug fix, refreshing content from an opportunity now opens the right tool.
FAQ
What is Prompt Intelligence in Hi, Moose?
It is a report on a single tracked prompt in Visibility. Moose reads every observed answer to that prompt, extracts the claims, groups them into recurring claims and answer patterns, and compares the AI systems against each other. Every finding links to the answers it came from.
Which plans include Prompt Intelligence and the Opportunities tab?
Prompt Intelligence is available on managed plans. The Opportunities tab works on every plan, and on managed plans it attaches Prompt Intelligence findings to each row as evidence.
What does a row in the Opportunities tab give me?
A job written as the action, what the answers show, the next step, the expected outcome, a priority with its reasons, a confidence rating, and an effort rating. Each row has a button that starts the work, such as Create brief, Audit page or Draft a pitch.
Why are AEO audits quicker now?
The small decisions inside an audit run on a fast decision model, each recommendation is pre-screened before the full review, and the remaining reviews run in parallel on GPT-6 Luna. This is on for paid managed plans and BYOK.
The one who's under the desk for all of it.
