Kevin Indig recently settled a controversy that has been running for half of the time since the introduction of the AI Overview. Does topic authority that rewards brands for topic depth rather than breadth, actually carry over to search AI? Or does ChatGPT have an empty slide where the generalist type can step in and hold a cold section?
Indig gained access to six months (January-June 2026) of ChatGPT response data from Semrush, covering 1,094 US categories, five prompts per category, over 50,000 products, and over 600,000 quotes. He published a breakdown of his Growth Memo Substack on July 20, with the article available on LinkedIn the same day. His answer is that subject authority is important for AI search, and the reason is robustness. Once a product has achieved the largest share of mentions in a category, it tends to stick to it.
That’s the clean version. The messier and more useful version, which hides several sections in his post, is that there is almost none on the market that is stable at the moment.
The Most Important Categories Are The Ones That No One Has
Indig builds three buckets. A category has a clear winner when one product appears in at least four of the five test orders and beats the runner-up by five percent or more. It is an emerging leader when the product leads at least three points without hitting that gap. Everything else, which means that no type leads to even three out of five information, is counted as uncorrected.
Apply those definitions to the full dataset, and June 2026 looks pretty wide open. Only 15.2% of the categories had a clear owner. Just over half, 53.7%, were open platforms with many credible competitors and no product close to locking the door.
Indig ranked all 1,094 categories by estimated AI search volume and divided them into two equal groups. The high volume segment accounts for 98% of all AI search requests in the sample, and has the lowest number of owners of the two, 11.3% compared to 19% of the low volume segment. Combined that and 89.3% of estimated AI search demand resides within categories without a clear owner yet. The categories that cost the most money to be taken are, at present, limited.
I’ve spent the last three years watching SEOs debate whether topic authority is a real algorithmic force or a simple news agency telling clients to justify a content calendar. Indig data is the first thing I’ve seen that solves the question with a real number. Title authority is real in AI search, and brands still treat category focus as optional offering nine persona categories to anyone who bothers to appear consistently across all five types of content.
Owners It’s Hard to Topple a Monarchy, But Only When the Lead Is Wide
The second part of Indig’s analysis is where the strategy gets sharp. He tracked month-to-month leadership changes in the same categories and found that the clear owner held first place in 90.4% of the comparisons.
The changes that occurred were concentrated almost entirely in categories where the lead was already small. Divisions where the leader changed had an average lead of just 1.3 percent to change. The categories where the leader held had an average lead of twice that, 2.9 points. The growth trajectory has told Indig almost nothing on its own. He shows three examples where a leading brand has grown month after month and is still being outperformed, because a competitor has either grown faster or closed a small gap. The lesson is not that momentum is meaningless. A lead of one or two points is not a lead worth defending at this point, and a brand sitting in that range should consider itself to be fighting.
Indig also discovered something that undermines common assumptions about how identities are formed. Citations and brand mentions appear to be only weakly related, with a small negative correlation of -0.229. The most cited domain in the category rarely matched the most frequently mentioned product, matching only 20.8% of the time. But the most mentioned type was cited at least once 69.9% of the time. Being overquoted doesn’t automatically make the ChatGPT version you automatically access. That’s a meaningful distinction for anyone with a GEO strategy that starts and ends with “get more quotes.”
The owners shared some characteristics compared to the runners-up: 55.7% higher keyword search volume, 48.4% higher organic traffic, and 52.5% higher Semrush Authority Score. Indig is careful to call this correlation rather than a formula. A brand does not earn ChatGPT’s autoresponder by purchasing branded search volume. It finds it by being a source of training and recovery of ChatGPT continues when the model tries to answer all five types of question, description, comparison, alternative list, use case, and purchase decision.
How to Use This Before Closing the Window
This is where I would point to the next quarter of work, based on the Indig framework and my Voice of Voice approach to tracking what AI says.
- Choose your battlefield first, then measure it properly. List 10 to 20 categories where your brand needs to be an auto-responder. For one, track the five types of information used by Semrush, a description question, a comparison, a different list, a use case, and a purchase question, not just your brand name alone.
- Sort by lead size, not by “win” yet. A stage where you lead by 1.5 points is not a stage you win. Treat anything less than 3 points as contested and place real resources behind closing the remaining information gaps, as Indig’s data shows that is the exact distance where profit occurs.
- Stop chasing quotes as a whole strategy. Use quotes to earn your way through the feedback loop, then shift the budget to specific content types, comparison pages, validation points, third-party coverage, and clear positioning, which determines whether it’s a truly branded product when you’re in-house.
The GEO discussion in the past year has been fixed almost entirely on citation counting because citations are the easiest thing to throw on a dashboard. Indig data says quotes are a door, not a room. Employees who continue to prepare the door while competitors promote the ChatGPT name are actually saying out loud that they will lose the categories they thought they were winning.
The window defined by Indig will not stay open. Eighty-nine percent of the demand that remains in the unprepared sectors is an unusually open start, and the brands that move on to it in the next few areas, instead of taking this as a single branding report, are the ones that will still be the default answer when someone does this same research in 2027.
Additional resources:
Featured image: Roman Samborskii/Shutterstock
