- GBP category: not what’s technically accurate, but what’s driving discovery in this market right now. It’s not a small lever either, Whitespark’s 2026 survey ranks primary category as the single highest-weighted relevance factor in the entire local algorithm.
- City and service area: the biggest nearby city is rarely the right answer.
- Homepage intent: what people actually come for, not just what the business technically is.
- Internal pages for secondary markets: don’t make the homepage compete with itself.
- Content grounded in real customer language: pull from reviews, not generic copy.
- Prioritization matched to trusted local ranking factors, not to whatever is easiest to automate.
Success also has to mean something different. Metrics need to reflect business outcomes, GBP actions, calls, direction requests, conversions, not just traffic. If your system isn’t reporting on those, it’s optimizing for the wrong thing.
Use AI to scale analysis, not to replace decision-making. It’s excellent at finding patterns across thousands of locations. It’s not going to tell you why one specific restaurant needs to target breakfast. That’s not a knock on the tools. OpenAI’s own research found that models are trained and graded in ways that reward a confident guess over admitting uncertainty, so a wrong answer often looks exactly as sure of itself as a right one.
The real problem is poor prompting and poor process design, not the tools. Tools execute whatever you design; if the design removes judgment at the wrong step, the tools execute that removal at scale.
Building systems that hold onto your judgment follows the same shape every time:
- Do it manually first
- Apply your why
- Automate only the clear logic
- Set a confidence threshold
- Validate against your own judgment on a regular cadence
That last step is the one people skip, and it’s the one that catches drift before a client does.
If you can’t evaluate a step independently, you can’t automate it responsibly. That’s the bar I judge it against, a test borrowed from a Wired series where physicists explain quantum physics to a child, a teenager, a college student, and a peer. If you can’t explain a decision at that range of levels, you have no business automating it, because you won’t see where it gets misunderstood until it already has.