AI Share of Voice: How to Measure It and Read It

AI share of voice is the share of AI answers that name your brand, compared with the competitors named in answers to the same questions. If 20 answers about your category name five brands, and yours appears in 6, that is one reading of your share. It tells you where you stand against alternatives, which a plain mention count does not.

Key Takeaways

• Share of voice is relative. The same mention count means different things in a crowded and an empty category.
• It depends on the question set. Use buyer questions that do not name any vendor, and keep them fixed.
• Measure across assistants. They answer differently, and an average hides which one leaves you out.
• A single run is an example. Trends need repeated readings over time.
• Being named is not being recommended, and neither is being cited. Record them separately.

What It Measures

For each question in a fixed panel, run it in each assistant and record which brands the answer names. Share of voice is your count divided by the total across you and the competitors you track.

Three details decide whether the number means anything.

1. The questions. Branded questions inflate the score, because you named yourself. Unbranded buyer questions ("what tools do X?") show whether an assistant volunteers you.
2. The competitor set. The share depends on who is in the denominator. Choose the brands a buyer actually considers.
3. The assistants. Lectern's standard reporting covers ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews. Hobby's weekly check skips Claude. Read the current coverage on pricing.

How Lectern Reports It

Lectern runs a defined set of buyer questions on a schedule, records what each assistant answers, and compares how often answers name your brand with how often they name the competitors you track. Hobby, the free plan, runs a weekly check of 10 buyer questions for one brand, and a new workspace that signs up with a work email gets $50 in welcome credits.

Lectern does not offer a signed-out share of voice reading today. You need an account to run one.

How to Read a Reading

• Look at the sample. A share built on 10 answers moves a lot with one answer. Check how many answers it rests on.
• Compare like with like. If the question set or the assistants change between runs, the change may be the panel, not the market.
• Split by assistant. A flat total can hide a gain in one assistant and a loss in another.
• Separate mention, citation and recommendation. An answer that lists you is not endorsing you. An answer that cites your domain may not name you.
• Expect noise. Assistants give different answers to the same question on different days. Repeated readings show a trend; one run does not.
• Do not claim cause. If your share rises after you ship something, that is an observation. Other things changed too.

What Moves It

Share of voice follows two things. The first is access: if an assistant cannot read your pages, it cannot use them. Check that with the free agent access check, read how to weigh its findings in the agent readiness score, and see which AI crawlers to allow. The second is evidence: pages that answer the question directly, and independent sources that say the same thing about you. AX vs AEO covers how these fit together.

Paid placements can add independent sources. They are labelled commercial content that complements earned coverage, not a replacement for it, and they do not guarantee that an assistant will name you.

Start Measuring

Pick 10 to 20 unbranded buyer questions and name your competitors. Sign up for Lectern's free plan to run them across assistants every week, or see pricing for daily runs and more brands. For the wider set of metrics around this one, see the AI visibility guide.