GEO assistant
Open the GEO assistant from the bottom-right corner to interpret results, research sources and move content forward in a side panel.
On this page
The GEO Assistant is a research assistant you can call up anywhere in the console. It reads your brand's profile, diagnosis results, answers and actions, searches and reads public web pages when needed, and replies with sourced findings and suggested next steps.
Opening the assistant
On any brand page, select the floating Open GEO Assistant button in the bottom-right corner. A persistent panel opens on the right:
- it doesn't cover or interrupt the page, so you can look at your data while you ask; on wide screens the main content shifts to sit alongside it;
- the conversation stays open as you move between pages of the same brand;
- conversations are saved automatically and listed under Recent conversations; choose New conversation to start a new topic.
The assistant is available on brand pages only; workspace pages don't show the button. Trial accounts don't include the assistant.
Three assistant types
Pick one under Assistant type:
| Type | What it's for |
|---|---|
| Brand research | Researches your website, competitors and positioning against your saved profile, separating confirmed information from missing or unverified details |
| Diagnosis analysis | Explains findings using saved answers, citations, audits and actions, and suggests next steps |
| Content optimization | Researches trusted sources for improvement actions and suggests an outline, usable sources and unsupported claims to avoid |
Each type offers an example question; choose Use example to get started quickly.
Attaching the current item
In Content, choose Ask the assistant about this draft to attach the current draft or topic. The panel shows Attached: current draft, and the assistant answers based on that version's body, FAQ and fact-check results. Choose Detach to remove it. The attached content is fixed for that question, so later edits to the draft don't affect the answer.
Reading the answers
Every finding states its basis:
| Label | Meaning |
|---|---|
| Observed | Backed by a source you can open under Evidence and sources |
| Inferred | Reasoned from available information, not directly confirmed by a source |
| Unverified | Not enough information; needs confirmation |
Sources are marked as Saved platform result, Read webpage or Search result. A search result is only a snippet, not a page that was actually read. Without diagnosis results, the assistant says so plainly rather than making up scores or conclusions.
Suggested next steps link to the relevant page, such as the brand profile, competitors, evidence, actions or retests.
From the assistant to content
An answer from Content optimization can be handed to Content with Write and fact-check: it creates a topic from an open improvement action, then generates a draft that goes through the independent fact check. The assistant's answer only guides the writing. Facts in the article must still come from the topic's sourced fact list, and the assistant's inferences never become verified facts automatically.
Verification actions can't produce content directly. Finish the verification and get an improvement action first; see Actions.
What the assistant doesn't do
- It doesn't change diagnosis metrics, the brand profile, action status or publication records. Every change still happens on the relevant page, with your confirmation.
- It doesn't publish content for you.
- Its answers are not a new diagnosis and don't count toward any metric.
Limits
- Each conversation allows up to 20 questions; start a new conversation after that.
- Each research round has a time limit. If it times out, narrow the question and ask again.
- A conversation handles one question at a time; follow up once the current round finishes.
- If a round is interrupted or fails, earlier turns are kept and you can Ask again.
Research runs in the background. If you close the panel or refresh the page, you can return to the conversation to see the result.