What Is Share of Model? Measuring Your Brand in AI Answers

By Rishabh Jha & Anuj Kamble, WisdmLabs · Updated 16 October 2026 · 5 min read

Share of Model (SoM) is how often AI answer engines such as ChatGPT, Gemini and Perplexity name your brand when people ask about your category, as a percentage of sampled answers. It’s the AI-search version of share of voice. There’s no standard formula yet, so pick one method and keep it fixed.

Most Share of Model numbers you’ll see come from tools that sell the measurement. You can measure brand visibility in AI answers yourself with a spreadsheet. But AI answers change from one run to the next, so a single check tells you almost nothing. Here’s a method that holds up.

Where does the term Share of Model come from?

The term was popularised by the marketing agency Jellyfish. In a 2025 INSEAD Knowledge article, an INSEAD professor and two Jellyfish staff describe it as a measure of how often, how prominently and how favourably brands appear in AI-generated responses.

The idea mirrors share of voice. Instead of counting rankings or ad impressions, you count how often an AI answer names you. That makes it part of answer engine optimization (AEO): the measuring half, not the doing half.

The engine you check matters a lot. In the same INSEAD article, Ariel held nearly 24% of laundry-detergent mentions on Meta’s Llama but under 1% on Google’s Gemini. Share of Model is a number per engine first, and an overall number second.

How do you calculate Share of Model?

The simplest version is a mention rate:

Share of Model = answers that name your brand
                 ÷ total answers sampled × 100

Illustrative example: 40 questions, run 3 times on each of 4 engines, gives 480 answers. If 120 of them name your brand, your Share of Model is 25%. Work out the same number for each competitor from the same 480 answers.

Tools mix and match these, which is why their numbers rarely agree. Only compare Share of Model numbers that were measured the same way.

How to build a prompt set that reflects real buyers

Your prompt set is the list of questions you’ll ask the AI tools. It decides what your number means, so build it carefully.

• Write 20 to 50 questions a buyer would ask, such as “best WordPress SEO plugin for WooCommerce”.
• Mix category questions, comparisons and problem questions (“how do I fix…”).
• Leave out questions that include your brand name. You’ll appear for your own name.
• Keep the set fixed, so this month’s number compares with last month’s.

Be honest about what the list is. As one tool builder put it on Hacker News, AI companies give “0 insight into what users are actually asking”.

So base your questions on real evidence: customer emails, sales calls, support tickets and the search terms in Search Console. A prompt set is an educated guess, so make it a well-sourced one.

How to measure brand visibility in AI answers, step by step

You don’t need a paid tool to start. A spreadsheet and an hour a month will do for one site.

1. Set up columns for date, engine, prompt, run number, brands named (in order), and pages cited as sources.
2. Pick the engines your buyers use. ChatGPT, Gemini, Perplexity and Microsoft Copilot are common starting points.
3. Run every prompt 3 to 5 times on each engine, in a fresh, logged-out session where you can, with the same location each time.
4. Log every brand each answer names, not just yours. Your competitors’ numbers come from the same answers.
5. Calculate Share of Model per engine and overall, for you and each competitor.
6. Repeat on a fixed schedule, such as the first week of each month.

Keep it clean: don’t paste client names, private data or unreleased plans into prompts. Use generic buyer questions only, and check your account’s data settings before you start.

Running this for clients? Use one prompt set per client and keep the same engines and run count across all of them. That’s what makes a monthly report comparable. A small, consistent sample beats a large one you can’t repeat.

Why does your Share of Model number jump around?

AI answers are not stable. In a 2026 study of 2,961 prompts by SparkToro and Gumshoe, ChatGPT and Google’s AI in Search returned the same brand list less than 1% of the time, and the same list in the same order under 0.1% of the time (Search Engine Journal). The study isn’t peer-reviewed, but the pattern is clear.

The same study found that how often a brand appears is far steadier than where it appears. Practitioners report the same thing: one commenter on Hacker News found that tracking top-three appearances over time was “much more stable than individual prompt outputs”.

• Personalisation. Account history, location and IP address can change answers, a point raised in a Hacker News launch thread for one tracking tool. Keep sessions and location fixed.
• Small samples. With 20 answers, one answer moves your score by 5 points. With 480, it moves by about 0.2.
• Mentions aren’t citations. Being named and having your page linked as a source are different wins. Track both.

Trust the trend across months, not the result of any single run.

What can you do to raise your Share of Model?

AI engines name brands they can find, understand and trust. Nobody can guarantee a mention, but these steps make you easier to name:

• Answer your category’s questions directly on your own site, in the first lines of the page.
• Keep comparison and pricing pages current, so engines quote today’s facts.
• Earn mentions on sites the engines already cite for your category. Your spreadsheet’s “pages cited” column shows you which ones.
• Let AI search crawlers read your site. See Block AI crawlers on WordPress for which ones matter.

A distinctive brand name helps too. As one SEO noted on Hacker News, a unique name makes it easier for people and models to attribute a mention to you.

To check that your pages answer questions clearly, Ranklance’s free plugin shows an AEO score for each page in the WordPress editor. It’s Ranklance’s own 100-point rubric, not a metric AI engines use.

Pair your Share of Model with the clicks you get: here’s how to track AI traffic in GA4. Share of Model shows if you’re named; GA4 shows if anyone clicks.

Share of Model worksheet: a 7-point checklist

1. Write 20 to 50 unbranded buyer questions from real customer evidence.
2. Choose your engines and write them down.
3. Decide your counting method (mention rate, top three or weighted) and stick to it.
4. Run each prompt 3 to 5 times per engine in a clean session.
5. Log every brand named and every page cited.
6. Calculate Share of Model per engine, for you and your competitors.
7. Repeat monthly with the same set, and compare trends, not single runs.

FAQ