Share of Model Reporting: Add SoM to Quarterly Dashboards

Your CFO asks for a single number showing how AI assistants are recommending your product this quarter. You run five prompts in ChatGPT and get inconsistent lists. The right answer is not a single prompt, it is a repeatable metric you can trend and justify in board decks. The metric you need is share of model reporting, the percentage of AI answers in a defined prompt set that mention your brand versus competitors.

Share of model reporting moves AI visibility from anecdote to dashboard. Define a fixed prompt set, run each prompt multiple times across target assistants, and report your brand mentions as a share of all tracked brand mentions. Most brands see measurable SoM gains within 90 to 180 days after starting a coordinated GEO program, with continued growth as authority signals compound https://everything-pr.com/what-is-share-of-model-the-new-metric-replacing-share-of-voice. Single-run checks are noisy; sample size and repetition matter for credible reporting https://lemniscategrowth.com/blogs/ai-share-of-voice.html.

Key takeaways

• Share of model reporting measures your brand mentions in AI answers as a percentage of total brand mentions across a fixed prompt set.
• Run 30 to 100 prompts per quarter and repeat each prompt multiple times per assistant to reduce variance and build trend lines.
• Report three layers: headline SoM, engine breakdown, and prompt-category performance for clear executive context.
• Combine mentions, citations, and sentiment with weighted scoring to surface priority fixes, rather than treating raw mentions as final.

What is share of model reporting and why add it now?

Share of model reporting is the practice of measuring your brand's presence inside AI-generated answers and presenting that number as a quarterly KPI. In one sentence, it is the fraction of AI assistant responses in which your brand appears, divided by total brand mentions in those responses.

Why add it now? AI assistants are already a recommendation surface that often precedes a website visit, analyst report, or vendor shortlist. The metric is conceptually the AI analog of share of voice, but it captures presence in generative answers rather than impressions or search queries. Definitions align across sources: Share of Model measures frequency, prominence, and favorability of brand mentions when users ask category questions https://agilebrandguide.com/wiki/agentic-commerce/share-of-model-som, and multiple glossaries state it as the percentage of AI answers recommending a specific brand https://cdp.com/glossary/share-of-model.

Adoption urgency comes from two facts. First, the behavior exists now: assistants recommend vendors and influence early-stage discovery. Second, SoM is measurable and moves with coordinated effort. Real programs report measurable improvement within three to six months after deploying GEO tactics, so quarterly reporting captures both initial impact and early compounding returns https://everything-pr.com/what-is-share-of-model-the-new-metric-replacing-share-of-voice. That timeline makes SoM a natural addition to quarterly dashboards for demand and growth teams.

How do I implement share of model reporting in my quarterly dashboards?

Start with a defined prompt set, run repeated samples across target engines, and calculate mention share by brand. That single sentence gives the workflow. Now here are the concrete steps, tools, and tradeoffs.

Step 1, pick prompts and engines

Choose 30 to 100 prompts that reflect buyer intent. Split them into categories: category definition, product comparison, recommendation, and reputation checks. Include organic prompts that do not mention brands and branded prompts for knowledge depth. Run prompts across engines you care about, typically ChatGPT, Perplexity, Gemini, and Claude.

Step 2, sample enough runs

Run each prompt multiple times per engine. The recommended minimum is three runs, but 10 to 30 runs per prompt produce more stable estimates for executive reporting. Variance between identical prompt runs commonly ranges from 10 to 25 percent, so modal or average results over many runs survive scrutiny https://lemniscategrowth.com/blogs/ai-share-of-voice.html.

Step 3, compute SoM and related metrics

The basic formula is simple: your brand mentions divided by total mentions of tracked brands, times 100. Use a modal or mean mention count across runs, not a single-run outcome. For engine-level SoM, compute per assistant and then produce an aggregate weighted by engine priority. Document your weighting and keep raw counts available for auditors.

Tools and tradeoffs

Manual runs are possible but scale poorly. Specialist AEO platforms and monitoring tools speed data collection and provide source traces. No single tool covers all engines and citation types yet, so expect to combine a purpose-built GEO tool with manual checks. For tools that integrate with dashboards, consult vendor reports on prompt sampling and engine coverage https://www.shadow.inc/resources/how-to-measure-ai-share-of-voice.

Which datasets feed SoM: mentions, citations, and sentiment?

SoM is primarily a mention share metric, but citations and sentiment materially change what the number means. Mentions tell you presence, citations show traceability, and sentiment shows favorability. Treat all three as separate signals in your reporting.

Mentions are the core input. Count named brand appearances in each valid AI answer. For consistency, specify what counts as a mention: brand name, product name, or common abbreviation. Count only mentions that are relevant to the user intent of the prompt set.

Citations capture when an assistant provides a source link or explicitly references a publisher. Citation share matters because many enterprise buyers and auditors treat source-backed recommendations differently. Track two citation metrics: citation frequency, and citation quality by source domain authority. When an assistant cites your content, that suggests higher trust and is worth bonus weighting.

Sentiment and framing capture tone. Not all mentions are equal. A brand mentioned as a top recommendation should score higher than one mentioned as a caveat. Use a simple sentiment rubric: positive recommendation, neutral mention, negative mention, and map that to weights, for example +1 for positive, 0.5 for neutral, 0 for negative. Combine weighted mention counts into a weighted SoM for a richer executive metric.

Tradeoffs include subjectivity in sentiment scoring and inconsistent citation behavior across engines. Keep raw data exportable so analysts can rerun weightings for different audiences or audit purposes.

Dashboard templates and KPIs for quarterly SoM reporting

Present SoM on three reporting layers: headline, engine breakdown, and prompt-category performance. That structure gives executives a single number, then the context needed to diagnose changes.

Headline panel

Show aggregate SoM for the quarter, period-over-period delta, and competitor ranking. Display the headline as a percentage with delta arrows and confidence intervals computed from sample variance. Include raw counts below the number, for example: 22.5 percent SoM, up 4.3 points, based on 3,600 answer samples.

Engine breakdown

Plot SoM per assistant. Use stacked bars or small multiples for ChatGPT, Perplexity, Gemini, and Claude. Engines behave differently, so call out the primary engine your customers use. Annotate spikes with model release dates or major content pushes to link cause and effect. Provide links to the raw engine transcripts for auditability.

Prompt-category heatmap

Show performance by prompt category: definition, comparison, recommendation, reputation. Use a heatmap with cells showing SoM and average sentiment. This highlights where your brand wins category definitions but loses direct comparisons, for example.

Complementary KPIs

• Citation share, citation quality score, and weighted SoM using sentiment weights.
• Top contributing content pieces and domains that lead to citations.
• Change drivers table linking content publication, backlink events, or PR campaigns to SoM movement.

For layout inspiration and GEO tactics, see our guide on how GEO can transform visibility /blog/how-generative-engine-optimization-geo-can-transform-your-br and the post on tracking presence across assistants /blog/track-brand-presence-ai-assistants.

How to set alert triggers and action playbooks for falling SoM

Trigger rules should combine magnitude, velocity, and channel scope. A good answer is to alert when SoM drops by more than 3 percentage points month over month, or when engine-level SoM drops by more than 20 percent of its previous value. That sentence gives a practical starting rule you can tune.

Define three alert tiers. Tier 1, watch: small but consistent decline over two measurement windows, no immediate action beyond monthly review. Tier 2, urgent: a sudden drop greater than threshold, require root-cause analysis and tactical fixes within two weeks. Tier 3, critical: multi-engine decline plus negative sentiment uptick or loss of citation coverage, require cross-functional teardown and rapid PR, product, or content interventions.

Root-cause checklist

• Check prompt-level transcripts for wording changes or hallucinations.
• Verify whether competitors published momentum content or earned citations that explain the shift.
• Confirm model updates or engine policy changes that may affect citation patterns.

Action playbooks

Content patch: update high-intent pages and create concise, model-friendly summaries and schema. Source seeding: prioritize authoritative citations on publisher sites and ensure canonical content is structured. PR and earned media: target placements on sites that AI assistants frequently cite. Product fixes: adjust feature messaging if sentiment shows consistent negative framing.

Automate alerts using your monitoring tool or a lightweight rule engine tied to your dashboard. For tools that surface source signals and recommended actions, consult AEO vendors and PR platforms that integrate citation tracking https://blog.hubspot.com/marketing/share-of-voice-tools.

FAQ

What is a good share of model percentage?

A good SoM depends on market concentration and category size. For competitive categories, a headline SoM above 25 percent is strong, while niche brands may target 10 to 15 percent. Compare your SoM to direct competitors and track movement, not just absolute numbers. Use engine-level benchmarks rather than a single universal target.

How often should I measure share of model?

Monthly measurement is the minimum for trend detection, while weekly checks make sense during active campaigns. Quarterly reporting is ideal for executive decks since meaningful change often shows in three-month windows. Continuous monitoring tools can alert you to abrupt shifts between scheduled measurements.

Can I calculate SoM manually without a tool?

Yes. Manual calculation requires running a controlled prompt set across chosen assistants, repeating runs per prompt, and tallying mentions. It scales poorly beyond a few dozen prompts. For credible quarterly reporting and auditability, use a mix of automated collection and manual spot checks.

Which AI assistants should I include in SoM reporting?

Include the assistants your customers use. Most teams start with ChatGPT, Gemini, Perplexity, and Claude. Add niche engines where you see buyer adoption. Report per-engine SoM, since models name different brands with different frequency and tone.

Does SoM replace SEO and other brand metrics?

No. SoM complements SEO, brand searches, and paid metrics. It captures early-stage AI-driven recommendations that can precede clicks. Use SoM alongside organic traffic, brand search share, and conversion funnel metrics to form a complete picture of discovery and demand.

Next step: build a 30 to 50 prompt audit this quarter and run repeated samples across two priority assistants. Put the headline SoM and engine breakdown into your next quarterly deck, and document sampling rules for the finance team. If you want automated data collection, traceable transcripts, and prioritized recommendations, consider tools that specialize in AI visibility; our platform at idou.ai focuses on measuring and improving brand presence across ChatGPT, Perplexity, Gemini, and Claude.