ModelSaid Agency Pro: automated AEO pull requests explained

ModelSaid Agency Pro adds a controlled implementation workflow to the Agency plan: connect a GitHub repository, analyze an eligible AEO issue, create a bounded change on a new branch, validate it, and open a pull request for normal review. It never pushes directly to the default branch. This is automation with a review boundary, not an autonomous website rewrite. Current prices and billing choices belong on pricing, where Stripe-backed terms stay current.

What Agency Pro includes before Auto Fix

Agency Pro is a superset of Agency. It includes ten business slots before purchasable extras, monthly visibility rescans, daily technical monitoring, the full searched-and-knowledge answer matrix, twelve-month trends, white-label reporting, thirty custom prompts, and fifty alert rules per business. Organization pools include fifty content drafts and ten Google connections. The plan expands collaboration to fifteen seats and on-demand capacity to thirty scans per month.

How an automated AEO pull request works

1. Connect the client repository through the Agency Pro GitHub integration.
2. Select an eligible issue where a bounded code change can address a documented gap.
3. ModelSaid analyzes the relevant repository context and proposes a narrow implementation.
4. The change is created on a new branch and validated before delivery.
5. A pull request is opened for the client's ordinary review and approval process.
6. Humans inspect, request changes, approve or reject; the default branch is never pushed automatically.

Why the branch and review boundary matters

AEO touches public claims, crawl rules and structured data, all of which can create real business risk if changed carelessly. A pull request exposes the diff, preserves repository controls and gives developers a familiar place to run additional checks. "Validated" does not mean "guaranteed correct for every business." The client remains responsible for factual truth, legal review, deployment policy and production approval. The boundary makes automation auditable rather than invisible.

Understand the Auto Fix allowance

Agency Pro includes up to two Auto Fix runs per UTC day and eight per month. Both limits apply: unused daily capacity does not convert the monthly ceiling into unlimited runs. Treat each run as a prioritized implementation slot. Batch related findings where appropriate, reject low-value cosmetic changes, and reserve capacity for issues whose expected benefit and evidence are clear. The AI readiness checker can help identify technical candidates before a repository is connected.

• Good: a missing or inconsistent machine-readable file with a clear desired policy.
• Good: structured data that can be aligned with verified visible business facts.
• Good: a narrow metadata or discoverability issue with objective validation.
• Poor: inventing testimonials, awards, locations, capabilities or unsupported "best" claims.
• Poor: broad redesigns, strategic positioning changes or edits requiring unavailable legal judgment.
• Poor: any change the client cannot review, test and own after merge.

Automation is most valuable after an agency has standardized diagnosis and approval. Define who may start a run, which repositories are in scope, who reviews facts, and what checks must pass before merge. Connect findings to client tickets and report the result of deployment in the next monitoring cycle. Use the schema generator to understand the visible facts a structured-data change should represent, even when Auto Fix prepares the implementation.

Who should buy Agency Pro

Agency Pro fits teams with several client businesses, established GitHub review practices and enough recurring technical AEO work to value bounded implementation. Agency is the better fit when clients want monitoring and white-label advice but do not grant repository access. Monitor suits a single business that does not need automated code changes. Test the demand with the free visibility scan, estimate how many evidenced fixes reach engineering each month, and compare that with the eight-run monthly allowance. The purchase case is shorter time from finding to reviewed change, not a promise of higher scores or automatic recommendations. ModelSaid monitors its own user-supplied #1 and near-100/100 positioning as a product practice, without presenting those claims as third-party awards. Review lead time from finding to merged change, pull-request rejection rate, and the share of runs that solve a documented client priority. Those measures reveal whether automation is reducing delivery friction or merely creating more review work.