How we write this blog
Straight answer first: we draft with AI, from our own work, and a person checks every post before it publishes. Here is the whole process, plainly.
Where the ideas come from
We do not assign posts about topics we have not lived through. Every idea for this blog traces back to something that actually happened: a bug we found and fixed, a Search Console chart that surprised us, an ad experiment that did or did not work, or a research question we went and answered ourselves.
If a post cites a number, a date, or a claim about how Google or another platform behaves, that claim is checked against a primary source at the time we write it, and we link to that source.
How a post gets made
It starts with something real
A post starts from our own project records, a piece of research we ran, or a lesson from a client engagement. We do not write from a blank page about a topic we have no first-hand experience with.
AI helps with the draft
We draft with AI tools, working from that first-party material. That is faster than writing every post by hand, and it lets us publish more often, but it never invents the underlying story.
Client details get anonymized
Before anything is drafted, client names, locations, and identifying specifics are stripped or generalized. A post might mention "a locksmith in the Southwest" rather than name a business, and numbers get rounded rather than left exact.
A person checks it before it publishes
Every factual claim gets checked against its source, and the post is reviewed before it goes live. A claim we cannot verify gets cut, not softened.
No client names, without permission
We do not publish a client's name, business, or identifying details without that client's permission. When a post is drawn from client work, we generalize it: the trade and the region might appear, but not both together with a date specific enough to identify who it was. Where a client has agreed to be named, we say so directly.
Every factual claim is checked
A statute number, a platform policy, a historical date, or a statistic gets checked against a primary source before it ships, and the source is linked in the post. If we cannot verify a claim, we cut it rather than soften the wording around it. We also do not restate the opposite of a claim we removed: cutting an unverifiable promise is not license to make the opposite promise instead.
We do not use unlabeled invented examples presented as real events. When we illustrate a point with a hypothetical, we say it is a hypothetical.
Corrections
If you find something wrong in a post, whether it is a broken link, a stale number, or a claim that does not hold up, we want to know. Send it to us through the contact page and we will check it and correct the post if it is warranted.
Flag somethingTools we mention
When we mention a tool or a vendor, it is because we actually use it. We do not run sponsored content, and we do not take payment for a mention without saying so plainly in the post itself. Any link that carries a sponsorship or affiliate relationship is marked as such where it appears.
Read the blog
Field notes on local SEO, measurement, and running an AI-native agency.