How I use AI in this work

Where does AI actually touch the work?

AI runs the measurement, sorts the data, reads your crawl and your Search Console exports, and produces first drafts I then rewrite. It also checks my own claims against their original sources before anything ships. Every one of those is a step where a machine reads more rows than a person reasonably can, and then a person decides what it means.

1. 01

   Runs the measurement

   The core of what I sell is finding out how ChatGPT, Perplexity, Google’s AI Overviews, and Claude actually answer your buyers’ questions. The only way to know that is to ask them, at volume, on a schedule. Your audit is a real set of buyer prompts run live through those systems, and your tracker re-runs them so you can watch the answers move. This is not AI writing about AI. It is measurement, and the models are the instrument.

2. 02

   Pulls and sorts the data

   Keyword demand, difficulty, ranking positions, and competitor footprints come from Semrush. AI helps me filter a few thousand rows down to the set that is actually winnable and worth building for.

3. 03

   Reads your site the way a crawler does

   I crawl your site with Screaming Frog, which is a standard technical SEO tool and not an AI. The crawl is the evidence: every URL, status code, title, canonical, redirect chain, and internal link on your site. What AI does is read that export and turn a few thousand rows into the short list of things worth fixing, in priority order. Same findings a careful human would reach, reached in an afternoon instead of a week.

4. 04

   Finds the patterns in your Search Console data

   I export your Google Search Console performance data, which is queries, pages, impressions, clicks, and average positions, and analyze it with AI. That is how I find the pages quietly decaying, the queries sitting at position 11 that are one improvement from page one, and the terms you already rank for and did not know about.

5. 05

   Produces first drafts

   I write from a brief, a voice profile, and a source list that I build. AI helps me get from that to a draft faster than typing it cold. Then I rewrite it. What ships is what I would have written anyway, sooner.

6. 06

   Checks my own claims

   Every statistic in anything I publish for you gets traced back to the primary source before it goes live, and the check is automated so it happens every time instead of when I remember. If a claim cannot be verified, it comes out. This is the step most AI content workflows skip, and it is the one that decides whether your content survives being read closely.

Where does your data go?

Into paid API accounts at Anthropic, OpenAI, Google, and Perplexity, plus Semrush for search data, and only ever your public content, your keyword and ranking data, your own crawl export, and your Search Console performance export. Your credentials, customer personal data, unreleased financials, and anything under NDA stay out unless we have agreed otherwise in writing.

Two of those are yours rather than public, so to be specific: a crawl export is a machine-readable list of pages you already publish, and a Search Console export is aggregated search performance, meaning queries, pages, impressions, clicks, and positions. Neither contains customer records or personal data. If your policy treats search performance data as confidential anyway, say so and I will work from summary figures I produce outside a model instead. It costs a little speed and nothing else.

Which systems, and why the API matters

I use paid API access from Anthropic, OpenAI, Google, and Perplexity, plus Semrush for search data. API access is a different privacy posture than the consumer chat apps, and it is deliberate:

• Anthropic states that by default it does not use inputs or outputs from its commercial products, including the API, to train its models (Anthropic Privacy Center).
• OpenAI states that data sent to its API is not used to train or improve its models unless you explicitly opt in, and that abuse-monitoring logs are retained for up to 30 days unless a longer period is legally required (OpenAI API data controls).

I have not opted in to data sharing on any of these accounts. Vendor policies change, so if this matters to your security review, ask me for the current position at the time you are reviewing and I will re-check the source rather than repeat what I wrote here.

Credentials

Any access you give me (analytics, Search Console, CMS) stays with me and gets revoked at the end of the engagement. You can revoke it yourself at any time without asking me. The logins themselves are never pasted into a model. The data I export using them is analyzed as described above, which is the entire reason for having the access.

How can you check my work?

• Ask me to show you the source for any claim in anything I write for you. There is one, and I will send it.
• Ask me to walk you through a prompt run live, so you can watch how the measurement actually happens.
• Ask for the raw exports. The Screaming Frog crawl and the Search Console pull behind any recommendation are yours.
• Look at the scoreboard. Your baseline is recorded before any work starts and re-measured on a schedule, so the results are checkable rather than narrated.
• Read anything I publish under my own name. I hold my own site and my own content to the same standard, and I publish the before-and-after numbers whether or not they flatter me.