Home Case study

GEO by Mod Op

Turning a manual AI search audit into a product clients could try for themselves.

Hand-drawn person scanning a landscape of changing signals with a magnifying glass

The client question

In early 2026, Mod Op's clients were asking how their brands appeared when someone asked an AI assistant for a recommendation. They wanted to know whether they were mentioned, which competitors appeared alongside them, and what they could do about it.

Our search and strategy teams had a manual audit for those questions. It took weeks to assemble, and covering more clients or more buyer perspectives would have meant more staff time. I worked with them to turn it into GEO by Mod Op: a free automated audit, with a paid program for clients who want deeper analysis and help acting on it.

What I owned

I led the product from a February 2026 hackathon prototype to a live release on June 18, 2026. I wrote the MVP specification, directed design and engineering, and developed the reporting framework and the commercial offer with our search and strategy experts, who supplied the domain expertise behind the audit.

Starting with the manual audit

The first prototype took about four hours during an internal hackathon in February 2026, one of a series of prototyping sprints I ran that spring. (The Launchpad case study covers the program.) Colleagues responded well to it in the prototype directory, and it was one of three working prototypes we showed a beverage client. Some clients saw AI visibility as a question none of their existing vendors had taken responsibility for. Those conversations moved GEO to the front of the roadmap.

We then had to decide what an automated audit should cover. A single prompt could not stand in for every buyer's experience, so the audit would test a set of likely buyer questions. We also wanted to examine the brand's website, since it is the material AI systems read and cite when they describe a brand. The MVP combined the two.

What the report shows

The audit starts with research into the brand and its category, then tests likely buyer questions across ChatGPT, Claude, and Perplexity. The report shows where the brand appears in those responses, how often it is cited, and which competitors are named.

We included the prompts, answers, and citations behind the score so a team could examine the findings themselves. A missing mention or an unexpected source gives someone a specific thing to investigate. The report describes the responses the audit sampled, not every answer a buyer might receive.

The framework

We organized the website assessment into the GEO Periodic Table, a framework of 25 factors covering content quality, structured data, brand and entity signals, trustworthiness, performance, and engagement. The system gathers evidence for each factor and assesses the site against it.

The results page explains how each factor was assessed and identifies potential improvements for the team to investigate. The framework gives our search experts and clients a common way to review the findings. It does not promise that changing a factor will produce a particular answer from an AI engine.

Deciding what to give away

We spent a lot of time on the boundary between the free audit and the paid program. The free report had to be useful on its own: an assessment someone could take to their team even if they never hired us. The paid program adds a deeper baseline, a cross-channel plan, and ongoing measurement with the agency's search experts, so its job is to help a client investigate the findings and make changes over time.

The audit went live on June 18, 2026.

Hand-drawn person tending a growing tree of signals and ideas

What I learned

I ended up favoring a more generous free report than I first expected. My reasoning was that a clear read on their position would give a team a reason to trust the people who produced it, and that the paid program would be easier to explain once the free audit had done its work. Whether that holds up is a question for the numbers Mod Op collects from here. It was the right call for the product we could ship in June.

Translating a manual process into an automation also showed me how much of the original practice was doing quiet work. The weeks our experts spent on an audit included moments to align with the client, ask a question, and apply judgment, and an automated pipeline skips those affordances unless you design them back in. The balance between agent-driven and human-driven steps is still a product decision, and getting it right is what lets you optimize for quality and speed at the same time.