Sign inStart free

The agents

14 agents. One job each.

Each agent owns one step of the loop. Small jobs are easier to check, and no agent checks its own work. Where a step puts your brand in public, you get a warning and a chance to check it before it goes live.

Together they do one thing: the work that gets your business named and cited by AI answer engines.

Start freeGet a free scan

Model-driven agents

Ten agents use a language model with fixed instructions: intake, discovery, answer analyst, gap diagnosis, content planner, writer, editor, listing scout, outreach and the copilot. Their output is checked by rules, by another agent, or by you.

Deterministic agents

Three agents use no model at all: the publisher, the experiment agent and the lifecycle agent. They follow fixed code, so the same input always gives the same result. Posting, statistics and alerts should not depend on a model's mood.

Understand

Learn the business and the questions buyers ask.

Intake agent

Model-driven

Reads your website and brochures, and builds the fact sheet every other agent works from.

Reads
Your website text and uploaded brochures, decks and PDFs.
Produces
The fact sheet, with each fact marked verified, VERIFY or RESOLVE.
Needs your input
YesYou confirm VERIFY and RESOLVE items.

Discovery agent

Model-driven

Builds the buyer questions that decide who gets recommended, grouped by what the buyer is doing.

Reads
The fact sheet, your competitors, countries and languages.
Produces
Blind buyer questions in groups, each with a tier.
Needs your input
YesYou can edit, add or drop questions.

Measure and diagnose

Read the answers and explain the gaps.

Answer analyst

Model-driven

Reads each engine answer: who was named, in what order, what was cited, and whether anything said about you is wrong.

Reads
Each engine answer and your fact sheet.
Produces
Who was named, in what order, what was cited, and any wrong facts about you.
Needs your input
Not neededNo. It only reads.

Gap diagnosis agent

Model-driven

Explains why you are missing from each answer and ranks the fixes by impact.

Reads
Many answers, their sources and the competitors named.
Produces
Gaps with one root cause each, ranked by size, with evidence.
Needs your input
Not neededNo. It recommends.

Fix on your site

Plan, write, check and publish pages.

Content planner

Model-driven

Turns gaps and discovery into a master plan of pages, with page types, targets, outlines and internal links.

Reads
Gaps, discovery results and the fact sheet.
Produces
The master plan: pages with type, target question, outline and ranking note.
Needs your input
YesYou choose which pages to write.

Writer agent

Model-driven

Writes each planned page in the citation-first format.

Reads
One planned page and the fact sheet.
Produces
A citation-first HTML draft, with anything we may not publish yet in hidden [VERIFY] comments.
Needs your input
Not neededGoes to the editor, then it is scheduled to go live on its own. You can read it, edit it or stop it first.

Editor agent

Model-driven

Checks every draft against the fact sheet and the rules before a human sees it. Never lets the writer grade itself.

Reads
The writer's draft and the fact sheet.
Produces
A list of issues (blocker, fix, note) and the VERIFY items you must confirm.
Needs your input
Not neededNo approval needed. A blocker still holds the page until it is fixed.

Publisher agent

Deterministic

Posts checked pages to your blog API or WordPress once their hold ends, then fixes internal links in a second pass.

Reads
Checked drafts and your publishing settings.
Produces
Live pages on your blog, then fixed internal links in a second pass.
Needs your input
Not neededRuns on its own once a draft's hold ends, unless you stop it first.

Fix off your site

Get onto the pages engines already read.

Listing scout

Model-driven

Finds the third-party pages engines cite for your questions, and how to get onto them.

Reads
The domains engines cited for your questions, with counts.
Produces
Each domain sorted by kind, with the honest way to appear there.
Needs your input
Not neededNo. It suggests.

Outreach agent

Model-driven

Drafts short, specific pitches to editors and platforms. It goes out automatically once its hold ends, unless you stop it first.

Reads
A target page, the gap in it, and your fact sheet.
Produces
A short pitch email under 150 words.
Needs your input
Not neededNo. It sends on its own once the hold ends, unless you stop it first.

Prove and keep going

Show what worked and what needs attention.

Experiment agent

Deterministic

Proves whether a change worked by comparing the questions you targeted against the ones you did not.

Reads
Checks on the questions we worked on and the ones we left alone, before and after.
Produces
A verdict: real improvement, real decline, or not distinguishable from noise.
Needs your input
Not neededNo. It only measures.

Lifecycle agent

Deterministic

Watches for drops, stale pages, low credits and listings or questions waiting on you, and tells you what to do next.

Reads
Your rates, page ages, credit balance and schedule queue.
Produces
Alerts and the next thing to do.
Needs your input
Not neededNo. It only notifies.

Help

Ask questions and run anything from chat.

GEO copilot

Model-driven

The chat assistant. Explains your results and runs any step for you when you ask.

Reads
Your data in the app, through tools.
Produces
Plain answers, and runs of any agent when you ask.
Needs your input
Not neededNo. The copilot never approves, publishes or sends anything itself. For listings and questions it tells you where to click.
How the stages connectStart free