Share of answer: the metric replacing rank tracking

Rank tracking answered one question: on this keyword, where do we sit in a list of ten links a person will scroll through? That question mattered when the search result was a list and the click was the prize. It matters much less now.

When someone asks ChatGPT, Gemini, or Google's AI Overview for "the best cold brew concentrate for a small cafe," they don't get ten links. They get a paragraph that names two or three brands, describes them, and moves on. There is no page two. There is barely a page one. The list collapsed into a recommendation, and a recommendation only has room for a handful of names.

If you are still measuring position 4 versus position 7, you are grading a test the market stopped giving.

What share of answer actually measures

Share of answer is simple to define and slightly annoying to measure, which is exactly why most brands don't do it yet.

Take a set of buying questions your customer actually asks — not keywords, questions. "What's a good magnesium supplement for sleep?" "Best running shoe for flat feet under $150?" "Which project management tool works for a five-person agency?" Run each one through the AI engines your customers use. Record which brands get named.

Your share of answer is the percentage of those answers where your brand shows up. If you appear in 6 of 40 questions, you have a 15% share of answer on that query set. That's the number. It's a percentage, it's benchmarkable, and it moves.

You can slice it the way you'd slice anything useful:

• By engine. You might own ChatGPT and be invisible in Google's AI Overview, or vice versa. They pull from different sources and reason differently.
• By intent. You might get named for "best budget X" and never for "best premium X." That tells you how the models have positioned you.
• By position in the answer. Being named first, with a reason, is worth more than being the fourth name in a trailing "other options include" clause.

That last point is why raw presence isn't enough. A good measurement counts not just whether you appear but how — first mention, described with a specific strength, or a throwaway.

Why rank tracking can't see any of this

A rank tracker checks a keyword against a results page and reports a number. It has no concept of a synthesized answer, no concept of being named versus omitted, and no way to tell you why a competitor got the recommendation and you didn't.

The AI answer is generated fresh, reasons over sources, and phrases a judgment. "Position" isn't a coordinate anymore; it's whether the model decided you were worth mentioning. Those are different physics. A tool built for the old one reports confident numbers about a race that's no longer being run.

This is the shift behind the line we keep repeating: channels don't grow brands, systems do. Chasing rank on individual keywords is channel thinking. Building the conditions under which models consistently name you is system thinking.

What actually moves share of answer

Once you're measuring it, the work becomes obvious because you can see what the models reward. Three things move the number more than anything else.

1. Being the source, not just being mentioned in one. Models cite and synthesize from pages that make clear, checkable claims. A product page that states materials, dimensions, use case, and who it's for — in plain language — is far more quotable than one built entirely of adjectives. If a model can't extract a fact from your page, it can't repeat that fact to a buyer.

2. Corroboration across the web. Models trust a claim they see echoed in several independent places. If your brand is described the same way on your site, in a handful of reviews, in a comparison article, and in a forum thread, the model treats that description as reliable and repeats it. If your only description of yourself lives on your own homepage, you're a single unverified source.

3. Answering the actual question directly. A lot of brands rank for a topic but never answer the buying question on the page. The model wants a page that says "for flat feet, look for X and Y" — not a 2,000-word article that circles the topic and closes with a coupon. Structure the page around the decision the buyer is making.

None of this is exotic. It's the same discipline behind everything we build: every claim you make, a buyer — or a model — can check in one tab. Pages that pass that test get named. Pages that don't, don't.

How to start this week

You don't need a platform to begin. Take your 20 highest-intent buying questions. Run each through two or three AI engines. Write down, per question, who got named and whether you were among them. That single spreadsheet is a baseline you can beat, and it will tell you more about your real visibility than a year of rank exports.

If you'd rather see your baseline scored for you, that's what our scan does — it measures where you show up across a real query set and shows the competitors winning the answers you're losing. From there the work is a system, not a keyword chase: structure the pages, seed the corroboration, and re-score.

Rank tracking measured a world of lists and clicks. That world is thinning out. Share of answer measures the one we're in now — where the search result is a short recommendation, and the entire game is being one of the names in it.

Want the full picture of how this plays out for your category? Start with a scan, or read how we run it end to end on the ecommerce side.