Deviate Labs

Bad creative, or bad tracking in disguise? Before you cut an underperforming campaign, make sure your conversion data is telling the full story.

Most companies have no idea how often ChatGPT, Gemini, or Claude mention them. And even fewer know how they stack up against their competitors. So we built a tool that tracked every relevant buying question for our client. Each month, we reported on which searches they appeared in across each AI engine, and periodically benchmarked their performance against their biggest competitors. And along the way, we collaborated with our client on weekly website content creation and broader website updates designed for Generative Engine Optimization (GEO). In December, the client appeared in 19% of the answers we tracked. By July, that number was at 90%. (Today, it’s at 97%) Across 35 questions and three engines (105 possible appearances) our client appeared 90 times. Their two biggest competitors landed 39 and 16. On Claude, our client showed up in 34 of 35 prompts. Every experiment I've ever run starts the same way: define what winning looks like, break the process down into parts, then go build it. So if you can’t define what winning looks like, you can’t optimize for it. If you don’t know your share of AI answers, you’re just playing roulette. You can’t optimize what you don’t measure. So if you don't know your share of the AI answers, you don't know whether you're winning or quietly disappearing. (You can see the full breakdown, including how we built the tracker, linked in the first comment below.)

You want another $20,000 for paid. Your CFO wants to know what the last $50k returned. Can you trust your data enough to answer?

Before you cut the campaign, check the tracking. On most accounts we audit, at least one conversion is counted twice, or not at all.

Bad tracking costs you twice: a report you cannot trust, and an ad platform optimizing toward the wrong signals.

Do you trust your marketing data? Download our free 9-Point checklist to spot common tracking gaps before they distort your decisions.