# How Sunny Patel runs SEO as an agent loop that grades its own predictions

Sunny Patel lets an AI agent run his site's SEO as a loop. Each cycle reads Search Console, makes one change, logs a prediction, and scores the result 7 to 14 days later. The full log is public.

| | |
|---|---|
| **Who** | Sunny Patel, SEO and AI consultant, Reading, UK ([LinkedIn](https://www.linkedin.com/in/sunny-patel-co-uk)) |
| **Problem** | Deciding what to change on his own site and knowing whether it worked |
| **Stack** | AI agent, Google Search Console API, a public experiment log |
| **Confidence** | High for the system. Outcomes still pending, by his own account |
| **Last verified** | 2026-10-13 |

## The problem
A consultant's own site gets the least attention. He wanted changes made and measured without doing each one by hand.

## The blueprint
1. The agent reads his Google Search Console data through the API.
2. It forms one hypothesis, edits the site and logs a prediction.
3. It returns 7 to 14 days later and scores itself.
4. Iteration 1 (7 July): title and meta rewrites on five striking-distance pages. One logged prediction: Reading-page CTR from 0.35% to at least 5% within 14 days.
5. Iteration 3 (10 July): the agent replaced a generic checklist download on two top posts with a two-card offer, a £495 fixed-fee audit plus a free Website Grader, with event tracking.
6. Iteration 4 was a conversion sprint he directed. The log flags it as three changes at once, so attribution is page-level only.
7. Every claim is checkable on his public /proof/ page.

## Stated results
Only the starting figures above. He writes that he "will not claim wins I have not measured." Outcomes: [UNVERIFIED / pending].

## Caveats
- X handle [UNVERIFIED] (no X link found on his site).

## Sources
- https://sunnypatel.co.uk/blog/autonomous-seo-agent/
- https://sunnypatel.co.uk/proof/