What share of model is
Share of model quantifies the frequency with which your firm or practice is named, described, or linked in the responses that large language models produce for prompts relevant to your services. Unlike traditional search share of voice, which counts impressions or rankings on a results page, share of model counts citations inside an answer. When a prospective client asks ChatGPT or Perplexity for a healthcare attorney in Chicago or a HIPAA-compliant telehealth platform, the model returns a synthesized answer. If your name appears in that answer, you have captured a unit of share of model. If you are absent, you have zero share for that prompt.
The metric depends on the prompt set. A broad prompt such as "best law firms for mergers" yields a different denominator than a specific prompt such as "firms with experience in Hart-Scott-Rodino filings for middle-market deals." CSToday defines the prompt universe for each engagement by mapping the questions your actual prospects ask. That universe becomes the denominator. Your citations across that universe become the numerator. The ratio is your share of model.
Related concepts include citation-worthiness, which describes the qualities that make a source likely to be cited, and answer-capsule, the concise summary that models often lift verbatim. Both feed directly into share of model. So does entity recognition: if the model does not associate your brand with the relevant entities, such as practice areas, jurisdictions, or regulatory frameworks, you will not be retrieved in the first place.
Why it matters for AI visibility
Prospective clients increasingly start their search in an answer engine rather than a traditional search box. They ask a question. They receive an answer. They act on the names inside that answer. If your firm or practice is not cited, you are not in the consideration set. This is zero-click-search behavior at scale: the click never happens because the answer is complete.
For law firms, a single citation in a model answer for "employment lawyer for executive compensation disputes" can drive a high-value consultation. For healthcare practices, a citation for "HIPAA risk assessment provider in Texas" can fill a pipeline. The commercial value of each citation is high because the intent is explicit. The user has already described the problem. The model has already filtered the universe. Your appearance is a qualified referral.
Share of model also reveals competitive dynamics that traditional rank tracking misses. A competitor may rank below you in organic search but above you in model citations because their content is structured for retrieval-augmented generation, because their llms.txt file guides crawlers to the right pages, or because their schema-org markup makes their expertise machine-readable. The LLMO Readiness Score surfaces these gaps by measuring the technical and content prerequisites for citation. The Answer Visibility Meter tracks your actual citation rate across a curated prompt set over time.
Where people get it wrong
The most common error is conflating share of model with brand awareness. A well-known firm may have low share of model if its website blocks ai-crawler access, if its content lacks the structured data that retrieval systems expect, or if its expertise is buried in PDFs that crawler-rendering pipelines cannot parse. Conversely, a smaller practice with clean llms.txt configuration, precise answer-capsule summaries on every service page, and entity-rich schema can capture disproportionate share.
Another error is measuring the wrong prompt set. Generic prompts such as "best lawyer" produce noisy, low-intent answers. High-value prompts mirror the language of your intake forms: "attorney for physician practice acquisition in Florida," "vendor risk assessment for BAAs under HIPAA." If you measure the former, you optimize for vanity. If you measure the latter, you optimize for revenue.
A third error is treating share of model as static. Models update. Retrieval indexes refresh. Competitors publish. Your share drifts. The AI Crawler Access Check ensures your site remains discoverable by the retrieval crawlers that feed the models. The llms.txt Generator keeps your crawl directives current as you add or retire content. Without continuous measurement, a lead today becomes a gap tomorrow.
What to do about it
Start by defining the prompt universe that matches your intake. Map the questions prospects ask at each stage: problem awareness, solution evaluation, provider selection. Use that map as the denominator for share of model.
Next, audit your citation-worthiness. Every service page should lead with an answer-capsule: a 40 to 60 word summary that a model can lift. Wrap expertise signals in schema-org markup: PracticeArea, MedicalSpecialty, Jurisdiction, ProfessionalService. Publish an llms.txt file that points crawlers to your authoritative pages and away from marketing fluff. Verify crawler access with the AI Crawler Access Check.
Then measure. Run the LLMO Readiness Score to baseline your technical and structural readiness. Deploy the Answer Visibility Meter to track citations across your prompt universe on a weekly cadence. Watch for drift after model updates or site changes.
Finally, iterate. When a competitor appears in answers where you do not, compare their page structure, their entity coverage, their answer-capsule clarity. Close the specific gap. Re-measure. The discipline is measure first, fix second, publish third. Radical honesty means showing the move when your share changes and explaining why.
CSToday runs this loop for law firms, healthcare practices, and regulated businesses. We build the prompt map, instrument the site, and prove the method by scoring our own site with the same tools every night.