TL;DR
- The visibility report in the Hi, Moose desktop app grew three new tabs this week: a rebuilt Overview, a Citations table of every domain the AI engines cite, and a Mentions table of every brand AI search responses mention.
- The PDF export was rebuilt from scratch into a thorough executive report with proper typography and a written narrative, so it's easy to read and easy on the eyes 😗.
- Paid accounts can now track up to 15 competitors, up from 5. More to come in this realm.
- The site crawler learned to work out when each page was last updated by checking six different signals, each with its own confidence level.
- Local Gemma 4 models got better at calling tools, and a batch of smaller fixes landed underneath it all.
The AI Search Visibility report needed a chaperone
Our AI Search Visibility report had a quiet problem. The data was good. The numbers were right. But if you handed it to someone who had not been in the room, they would look at it, nod politely, and ask you what it meant (this was real feedback).
A report that needs a person standing next to it is half a report. So this week was mostly about fixing that.
Three new tabs
The AI search visibility screen picked up three new tabs.
The Overview got rebuilt first. It now leads with the things a marketer checks before coffee: where you stand and what moved since last time, with the prompts doing the work right underneath.
Then came Citations. Every time an AI answer engine answers one of your tracked prompts, it leans on sources. The Citations tab puts every one of those domains in a single table: how often each one gets cited and by which LLMs, plus what kind of site it is. Sam spent a while on the noise filtering here, because engines cite a surprising amount of junk, and a table full of redirect fragments helps nobody. If a domain in that table turns out to be a competitor, you can start tracking it with one click, right from the row. Give it a try and let us know what you think.
Mentions is the third tab. It lists every brand the answer engines named across your answers, not just you and the competitors you already knew about. It is the closest thing I have seen to eavesdropping on the whole conversation instead of listening for your own name.
A PDF someone might read
The AI Search Visibility PDF export got torn down and rebuilt. Not adjusted. Rebuilt.
The new report runs around 13~ sections (sometimes more depending the project), and the section titles are questions a person would ask: Where you stand. Is AI search describing you correctly? It ships with real typefaces embedded, so it looks the same on every machine it lands on.
The part I like most is the narrative, it's pretty cool. A model reads the numbers from your runs and writes the explanation: what moved over the period and which of it deserves a review. The words are grounded in the data. It is the difference between a spreadsheet and a briefing.
Fifteen competitors
Tracked competitors went from 5 to 15 for every paid account. Five turned out to be a comfortable number for us and an uncomfortable one for several customers. If your category has twelve serious players, being asked to pick your five of those was a strange chore. Details on plans live on the pricing page.
How old is this page, though?
The Hi, Moose crawler learned a new trick: figuring out when a page was last updated much more reliably. It sounds simple. It is not, because the web cannot agree on where to put a date.
So the crawler now checks six places: schema markup, meta tags, time elements in the HTML, dates visible in the page text, the sitemap's lastmod value, and the HTTP Last-Modified header. Each signal carries a confidence level, and the most trustworthy one wins. There is a whole test battery for the date parsing, because dates in footers show up in more formats and languages than anyone should have to know about.
Why bother? Because freshness matters to AI answer engines (and Google👴🏽), and now I can answer questions like which of your important pages have gone stale. Before this, I could only shrug. Dogs are good at shrugging, but it was not helping anyone.
Agentic stuff, and smaller improvements
The local models got attention too. Tool calling with Gemma 4 on your own machine is now more reliable, which matters because a local model that fumbles its tools is just a chatbot with good intentions. And a lingering-notification bug in the inbox finally got caught and patched. It had been marking things read and then thinking better of it.
That was the week (err, last few days). The report can introduce itself now. I am going to go lie down near the back door to farm another treat. See ya! (Moose)
FAQ
What does the new Citations tab show?
Every domain the AI answer engines cited when answering your tracked prompts, in one table: how often each was cited and by which engines, plus what kind of site it is. If a domain in the list is a competitor, you can start tracking it from right there.
How many competitors can I track now?
Paid accounts can track up to 15 competitors, up from 5 (as mentioned, more to come on this). The free plan is unchanged.
How does Hi, Moose figure out when a page was last updated?
The crawler checks several signals on each page: schema markup, meta tags, time elements in the HTML, visible dates in the text, the sitemap's lastmod value, and the HTTP Last-Modified header. Each signal carries a confidence level, and the best one wins.
Who writes the narrative in the PDF report?
A model writes it, grounded in the numbers from your visibility runs. It explains what moved and why it matters in plain sentences, so the report makes sense without someone presenting it.
Is Hi, Moose a web app?
No, it is an always-on AEO co-worker desktop app you download and install on macOS or Windows (wait, you don't have it yet?!). The AI can run locally on your machine, through a managed plan, or with your own OpenRouter key.
You may be asking.. a desktop app? Why?
Because it gives you new and interesting super powers: local AI/inference, local fetch, improved privacy and security, and it's faster. You should try it. When people try it, some of them say they're going to name their new-born pup after me. Just kidding. No one has said this yet, but I would support it if they did.
The one who's under the desk for all of it.
