# Testing SEO Agents on Our Own Website: The Plan

Canonical page: https://2bcome.ai/experiments/our-own-site/
Content updated: 2026-09-15

EXPERIMENT 001 Planned test

# Can our agents improve the website that sells them?

The first proposed experiment uses 2bcome.ai as a bounded test environment, with a recorded baseline and approval before publication.

By [2bcome](https://2bcome.ai/about/) Written September 15, 2026 Updated September 15, 2026

Planned experiment. No results are available. This page describes the proposed test, not a completed case study.

## Status

Planned. No baseline has been collected and no results are available. This page is a protocol, not a case study.

## The question

Can an SEO, Content, and Web agent workflow identify a useful change to a defined section of 2bcome.ai, produce an acceptable update, and leave enough evidence to evaluate the work?

The first question is whether the workflow produces useful, controllable work. Search performance is a later outcome to observe, not a result we can invent at the start.

## The proposed scope

Choose a small section of the site and record its current pages, content, and relevant baseline data. Exclude pricing, legal pages, credentials, and infrastructure changes.

The SEO role proposes one bounded improvement. The Content role prepares the draft. The Web role implements a reviewed change and runs the agreed checks.

Publication requires a human approval tied to the actual revision.

## What to record

Capture the original brief, the sources used, the agent outputs, the review decisions, the validation results, and the final revision.

Also record how much human intervention was required and where the workflow stopped or failed. A test that only works after unrecorded manual repairs should not be described as autonomous.

## What would count as a useful first result

A useful first result is a change that a reviewer accepts as relevant, accurate, and within scope, with a clear audit trail and a recoverable publication process.

Any later performance review should use the agreed baseline and observation window. It should distinguish an observed change from a claim that the agent caused it.

## When we will report an outcome

Only after the baseline, intervention, and review have actually happened. The update should include negative and inconclusive results as well as positive ones.

[Explore the proposed team](https://2bcome.ai/teams/organic-growth/) or read [how we intend to measure the work](https://2bcome.ai/blog/measuring-what-matters/).

CONNECTED WORK

## The next useful connection.

AGENT TEAM

### [AI SEO Team: Research, Content and Website Changes](https://2bcome.ai/teams/organic-growth/)

Explore a proposed AI SEO team combining search analysis, content preparation, and tested website changes. Shared goals, scoped handoffs, and human review.

[Explore](https://2bcome.ai/teams/organic-growth/)

FIELD NOTE

### [Building an AI Agent Company with Our Own Agents](https://2bcome.ai/blog/our-first-customer/)

A practical reason to start with SEO agents: the website that introduces them also needs their work.

[Explore](https://2bcome.ai/blog/our-first-customer/)

FIELD NOTE

### [How to Measure AI Agent Outcomes](https://2bcome.ai/blog/measuring-what-matters/)

How we intend to evaluate agent work without mistaking a busy log for a useful result.

[Explore](https://2bcome.ai/blog/measuring-what-matters/)