Engineering notes

Short, dated notes on AI search.

Practical checks and opinion pieces on how AI apps read and name businesses. Each note is dated. Opinion is labelled as opinion, and any number carries its source.

Open source ·

Open sourcing the tooling: claude-rank, Ultraship and codesight

Three Claude Code plugins, published free on npm under the HouseofMVPs publisher account, run the same kind of technical checks we use when we review a site.

Read the note→
Diagnostic ·

The curl test: a one minute check that your content is in the HTML crawlers receive

One command shows whether a page's key text is in the HTML a crawler receives, or only appears after JavaScript runs. It works on any site, whatever it is built with.

Read the note→
Market shift ·

How GLP-1 medicines changed the questions bariatric patients ask AI

Many patients who once would have started with a surgical consult now start on a GLP-1 medicine. The questions they later ask AI apps are different, and most surgical center sites do not answer them.

Read the note→
Diagnostic ·

Reading a robots.txt for AI crawler access: the user agents to check first

Open /robots.txt and check that nothing blocks the crawlers behind AI search citations. Training crawlers are a separate, licensing choice.

Read the note→
Vertical analysis ·

Why dental chains show up in AI answers, and what independent cosmetic practices can do

Large dental groups publish consistent location pages at scale. Independent cosmetic practices can compete by being specific about the work they do and the dentists who do it.

Read the note→
Legal vertical ·

Bar advertising rules and AI search: keep disclaimers next to results

AI apps quote selectively. A verdict figure can travel without the disclaimer that sat in a footer. Put the disclaimer in the same passage as the claim.

Read the note→