The Hi, Moose blog
Notes on AI search: how answer engines decide who to cite, and how to become a source they trust.
Featured · Build in public
The graph your site already wrote
AI assistants retrieve fragments of your site and infer who you are, what you offer, and how it all connects. Now you can see the graph your pages add up to, and take it with you as CSV or PDF.
Build in public
The question everyone asks first
Ask "Does AI recommend my business?" in chat and find out in a minute or two, across five engines, free, from your own computer.
Build in public
Your Logo on the Cover
A report with the client's agency on the cover is invoiceable in a way our logo never was. This week we made our own product quieter, on purpose.
Build in public
The Draft That Forgot Who You Were
You give each project a brand voice, and every draft is supposed to read it before writing a word. Drafts made by scheduled workflows did not. The quiet kind of bug, now fixed.
Build in public
The Run That Didn't Wait for Wifi
Close the laptop, catch a train, and the visibility schedule still came due. The run fired, asked its questions into a dead connection, and got nothing. Fixed: runs now wait for real internet.
Build in public
The Audio Feature Moved In
One feature stayed behind when the desktop app launched. This week it moved in: Listen now lives in the desktop app, and the Visibility dashboard learned to export a report you can hand to your boss.
Build in public
The doors are open
Hi, Moose is in open beta. The desktop app is up for download on macOS and Windows, and launch day came with the usual pile of small fixes, including a settings toggle on Windows that swore it was off while the feature was on.
Local-first
Meet the new Moose: your AEO coworker now lives on your laptop
I rebuilt Hi, Moose as a desktop app with a real AI model on your machine, on a bet that visibility tracking alone won't survive. Here's the argument, and what shipped.
Build in public
The local model picked up tools
Chat tools used to be a hosted-model thing. As of this build, Gemma 4 running on your own machine can use them too. Plus: updates that wait their turn, and an analytics toggle written the honest way.
Build in public
The chat learned to make pictures
The chat can generate images now, read your spreadsheets, and it stopped guessing when you ask what your site hasn't covered.
Build in public
The model was right there the whole time
Our site crawler was tripping cloud rate limits while a capable model sat loaded on the same machine. Now the local model takes the first pass, and finished audits explain themselves.
Build in public
Chat learned to fetch
Chat can now go get things instead of describing them: your site's own pages, your Search Console data. Also, our internal-linking feature had a bar so high nothing real could clear it.
Build in public
Webflow has opinions about fields
Webflow collections each have their own required fields. The app now asks about them before publishing, not after the rejection.
Build in public
One draft, pushed once
You could push the same draft to your WordPress site twice. The app now remembers, and says no.
Build in public
The Day the Audit Ignored Me
You typed "aeo audit". The app thought about it, and then explained what an AEO audit is. Reader, it did not run one.
Playbooks
Treating AEO like CRO: controlled experiments, repeatable tests, measurable deltas
Give an LLM a fixed set of pages and a query, ask which it would cite, then iterate and measure the delta - AEO run like a CRO experiment.
Deep dives
Citation Alignment explained: AEO and AI Mode grounding queries
Google exposes the grounding queries behind AI Mode answers. Citation Alignment uses that data to see how your content aligns with what Google cites.
Honest notes on AI search.
No growth-hacks, no spam. Just what we're seeing change across the answer engines, and what we'd do about it.