TL;DR
- Chat now picks the right tool from a plain sentence: audits, briefs and drafts, FAQs, key points, Search Console questions, web research, the entity graph, AI visibility checks, Google AI Mode grounding queries, ChatGPT fan-out queries, audio, and versus pages. You do not need to know what any of them are called, and it works in English, Spanish, French, German, Italian and Portuguese.
- The tool pick runs on a small, fast decision model and is on for every plan, including chat with a local model. Settings > Local AI has a Smart routing switch if you would rather keep that step on your machine too.
- When I suggest prompts to track in chat, each one now has an Add button. One click puts it on your Visibility schedule, and you can add a whole category or all of them at once.
- Settings has a new Account logins section. Sign in to ChatGPT, Google AI Mode, Grok, Perplexity, Reddit, LinkedIn and X there once, and every visibility run and Send Moose job from your machine uses those sessions.
- Managed visibility scoring now tries up to three models in turn with a time cap on each, so one slow reply does not leave a result unscored. GPT-6 Sol, GPT-6 Luna, Claude Opus 5.5 and Grok 4.7 joined the model picker.
Ask in your own words
Some folks may not know that the thing they want is called an entity graph. They say "show me how my pages connect" or "what does Google think my company is". Chat handles that now. When you send a message, a small decision model reads it and picks one of thirteen routes: an AEO audit, a content brief or a new draft, FAQs, key points, a question about your Search Console data, sending me out on web research, the entity graph, a check of how AI assistants answer about your brand, the grounding queries behind a Google AI Mode answer, the fan-out queries behind a ChatGPT answer, audio, versus pages, or a plain chat response.
A direct ask for a tool by name, of course opens that tool. A hint gets a suggestion chip instead, so "this page feels weak for AI answers" offers you an audit and "run an aeo audit on this" runs one. If you ask for a landing page or a glossary post, the brief card opens already set to that kind of page. Whatever language you write in, the meaning is what gets routed. The test set we run it against includes Spanish, French, German, Italian and Portuguese messages.
The decision is lightning-fast taking about a quarter of a second. When the model is not confident, the previous routing approach takes over. Moose (me) is one fast dog, dawg.
This is on for every plan, including when you chat with a local model IF, you want it on. Under Settings > Local AI there is a Smart routing switch. See the Settings page in the docs. With Smart routing switch turned off, the tool selection action stays on your machine as well.
Prompts you could track, one click away
Ask me which prompts you should be tracking for your brand and the answer comes back as a card with each prompt having an Add button. Add one, add a category, or add all of them, and from then on they run on your Visibility schedule and build the trend like any other tracked prompt. A suggested category becomes a new category in Visibility, or joins an existing one when the name matches.
Sign in once, in Settings
Visibility runs from your machine and 'Send Moose' jobs both use my own browser, and some sites want you signed in before they will render a response. Settings has a new Account logins section that lists them: ChatGPT, Google AI Mode, Grok and Perplexity for visibility checks, and Reddit, LinkedIn and X for Send Moose. Each one shows Signed in or Not signed in, and clicking it opens the site in my browser so you can sign in ahead of time. Every run after that picks the session up.
Scoring with a backup plan
On managed plans, each visibility answer is scored for sentiment and for which brands it mentions. That scoring now runs down a chain of up to three models, GPT-6 Luna first, with a ten-second cap on each attempt. A reply that stalls or comes back without a verdict hands off to the next model instead of leaving the field blank.
The model picker changed in this latest build v0.6.26. GPT-6 Sol, GPT-6 Luna, Claude Opus 5.5 and Grok 4.7 are now available, and GPT-6 Luna is the new default for quick jobs. A few older models we retired since they were not used as often and in some cases cost more tokens: GPT-5.4, GPT-5.4 Mini, GPT-5.4 Nano, GPT-5.5, Claude Opus 4.7 and 4.8, Claude Sonnet 4.6, Grok 4.3 and 4.5, and Muse Spark 1.2.
Also this week
ChatGPT and Grok checks that run from your machine got another reliability pass. On some networks, a page loaded over HTTP/3 (QUIC) can hang without ever reporting an error, which showed as a check that never completed. This is now resolved.
Also, a bug fix: asking me in chat to send Moose out on a research job now starts the job first, then tells you it is on its way.
The docs got a page for Brand documents, and the Settings and model selection pages were updated for Smart routing.
If the decision model keeps holding up in chat, the next place we want to try it is deciding which crawl changes are worth an alert. Nothing built there yet.
FAQ
Do I need to know the tool names to use chat in Hi, Moose?
No. Describe what you want in your own words and chat picks the tool: an audit, a brief, FAQs, key points, the entity graph, an AI visibility check, grounding or fan-out queries, audio, versus pages, web research, or a Search Console answer. A direct ask opens the tool, and a hint gets a suggestion chip.
Does smart routing work when I chat with a local model?
Yes. Smart routing is on by default on every plan, including chat with a local model. Settings > Local AI has a Smart routing switch, and with it off the tool pick stays on your machine too.
How do I add a suggested prompt to Visibility tracking from chat?
Ask which prompts you should track and the answer comes back as a card with an Add button on each prompt. Click Add on one, or add a whole category or all of them. They then run on your Visibility schedule, within the prompt slots on your plan.
Where do I sign in to ChatGPT or Grok so visibility runs from my machine work?
Settings > Account logins. It lists ChatGPT, Google AI Mode, Grok, Perplexity, Reddit, LinkedIn and X with a Signed in or Not signed in status. Click a site to open it in Moose's browser and sign in, and every later run uses that session.
The one who's under the desk for all of it.
