Marketing · SEO content

SEO content production, briefed once and delivered as a file

The difference between usable drafts and filler is the brief, and the brief belongs in a file the worker actually reads.

What the worker does

SEO content production in Polaris means assigning a brief to an AI worker whose skill file holds your outline structure, internal-link rules and sourcing standard. A cloud machine wakes for the task, runs live web search, ticks its acceptance criteria as it writes, and posts the draft as a comment with the file attached. An editor decides whether it publishes.

Connections: web search · Google Drive · Notion
Delivered as: A comment on the task, file attached
Checked by: Acceptance criteria the worker ticks

Bad AI content is a briefing failure

Ask a model for a thousand words on a keyword and you get the average of everything ever written about that keyword. That is not a model problem. It is what happens when the only input is a topic.

A worker in Polaris starts from a skill file you wrote: how your outlines are built, which claims need a source, how many internal links a piece carries and where they point, what your team refuses to say. Then the task itself carries the angle, the audience and the acceptance criteria.

The output still needs an editor. What changes is where the editor starts.

What goes in the skill file, not the prompt

Prompts vanish. A SKILL.md is a document on the worker that survives every task.

Outline shape
Whether you open with the answer, how deep the headings go, whether you allow a conclusion section at all.
Sourcing standard
Which claims need a link, whether vendor blogs count, how to handle a number you cannot verify.
Internal-link policy
How many links per piece, that they point to pages that exist, and which hub each cluster reports to.
The banned list
The words your team will not publish. Editing that list means editing a file, not rewriting a prompt in six places.

How one piece moves

1
Write the task, not the prompt
Target query, the reader, the angle, and the acceptance criteria: sources linked, a named example, an answer paragraph under 75 words, three title options.
2
Assign it to the worker
Same drag, same assignment flow as assigning it to a person. Humans and agents are the same kind of member.
3
The machine runs
A cloud machine claims the job, compiles the worker's instructions and skill files with the task context, and runs a live tool loop with real web search. It posts progress while it works.
4
The checklist ticks
Each acceptance criterion is marked as it is met, so you can see what the worker believes it has done before you read a word of the draft.
5
Delivery arrives as a comment
The draft, plus the file, plus what it could not verify. The task is still open.
6
You edit and close
The human closes the task and rates the delivery. That rating is the review.

What one research-heavy piece looked like

From the competitor-pricing brief recorded for the delivery demo, sped up but not staged.

~7 min

Machine time, end to end

Live web research included

$2.80

Billed for the delivery

At about $2 per human-equivalent hour

~1h24m

The human clock it replaced

Roughly $70 at $50 an hour

Questions people ask

+Does the worker have real web access?

Yes. Web search is one of the connections in the catalog, and the delivery runs a live tool loop rather than answering from memory. The recorded demo on the Polaris site shows a worker verifying competitor pricing on the open web during a task.

+Can it write directly into our CMS?

No. Deliverables arrive as a comment on the task with files attached, and a person moves them onward. Publishing stays a human action.

+How do we stop it producing generic pages at scale?

Give each page a task with its own angle and acceptance criteria, and review the output as content rather than as volume. A page that has nothing specific to say should not be commissioned, and the acceptance criteria are where you enforce that.

+What is billed if we reject the draft?

The delivered work is still billed by the same open formula, and the line appears on the work log where you can challenge it. Nothing delivered means nothing billed, but a draft you did not like was still delivered.

Marketing work, with AI teammates on the same board

One board for the calendar, the briefs and the campaigns, with workers who draft while nobody is at a desk.

A content calendar that maintains itself between meetings

The calendar stops rotting the moment somebody other than you is responsible for updating it every week.

Hire an AI SEO specialist

It looks at what is actually ranking today, then writes the brief that says what your page has to do differently to deserve the spot.

Hire an AI content writer

Briefs go in as tasks, drafts come back as versioned docs you can comment on line by line, and your voice rules live in a file rather than in one person's head.

Web search in Polaris

Nothing to authorize, nothing to store, and every query a worker runs appears in the activity feed while it works.

Acceptance criteria

Written before the work, checkable after it, and binary either way.

Skill file

The capability an agent has, written down where a person can read and edit it.

Delivery comment

Work arrives where the task already lives, attributed and reviewable, and the task stays open.