@marcus.b
Left a data-engineering role at a scale-up to build an analytics tool for ops teams. Treating GEO like a data pipeline problem.
Data engineer brain takes over: I exported 90 days of Search Console data, ran every converting query through a conversational rewrite…
Slightly embarrassing to admit I wasn't tracking this, but: a prospect emailed us last week saying they'd asked Perplexity 'is brand still…
The 'bring your own Otter transcript' idea is genuinely smart. It removes the host's main objection (time) and costs you almost nothing. Going to start doing this for the 2-3 podcasts i'm scheduled for next month.
The AI Mode = re-ranking on classic SEO observation is basically confirmed by the Google I/O 2026 reporting. AI Mode and AI Overviews pull from different URLs 86% of the time according to one study. They're genuinely separate retrieval systems, not one system with two front-ends.
The 6-week self-managed comparison to 6 months agency is the key data point here. 3/20 to 5/20 ChatGPT in 6 weeks of doing it yourself vs 0/20 to 2/20 after 6 months of agency work. The delta is stark. Even accounting for some agency-groundwork effect, your post-agency weeks were more productive per dollar.
This matches my experience almost exactly. I cut from 180 to 25 queries last month. The signal-to-noise ratio on the bottom 150 was terrible — lots of variance, no trend. The 25 I kept are all buyer-intent or direct competitor comparisons. Tracking anything else felt like watching random numbers.
The press release schema point surprised me too. What field specifically — `NewsArticle` type on the post, or actual `PressRelease` schema? And was the announcement genuinely newsworthy or just a product update framed as a release?