Share of model

Share of model is a practical observation metric for AI visibility work. It asks: across a fixed set of prompts, how often did a product visibly name your brand? That can be useful for tracking answer presence over time, but it is not the same as citation quality, referral traffic, or business impact.

Direct answer

If a team asks whether share of model is “the AI ranking metric,” the safer answer is no. Google says AI features in Search use the same SEO foundations as Search overall, require no special AI-only markup, and report traffic within Search Console’s overall Web search type; see AI features and your website and the AI optimization guide. OpenAI separately documents that OAI-SearchBot, GPTBot, and ChatGPT-User serve different purposes; see Overview of OpenAI crawlers.

Taken together, the defensible inference is narrow: share of model is a publisher-run comparison method, not an official cross-platform metric supplied by Google or OpenAI.

What current first-party guidance supports

• Google’s AI features documentation says there are no additional technical requirements to appear in AI Overviews or AI Mode beyond normal Search eligibility.
• Google’s AI features documentation also says traffic from AI features is included in Search Console within the overall Web search type, which helps with Google-side measurement but does not create a cross-platform share-of-model report.
• Google’s AI optimization guide says third-party tools do not have access to Google’s internal ranking or AI systems and warns against chasing inauthentic mentions or creating page variations just for AI wording.
• OpenAI’s bots documentation says OAI-SearchBot is for ChatGPT search results, GPTBot is for training-crawl controls, and ChatGPT-User can fetch pages in response to user actions. Those are different events and should not be merged into one visibility story.

That means a useful share-of-model program must stay explicit about scope, test design, and limits.

Page boundary

If you need to answer…	Better page
how often a brand is visibly named across a fixed prompt set	this page
whether the cited source actually supports the claim	AI citations
how often your page is cited rather than merely named	Citation rate
whether answer visibility produced visits and conversions	AI referral traffic
whether a platform can crawl, index, or fetch the page	AI crawlers, ChatGPT Search, or the SEO audit workflow

This page should stay focused on answer-level brand presence, not absorb every AI visibility question.

Build a repeatable measurement set

Use one fixed design per run:

1. Define a stable prompt set tied to a real topic, market, and buyer task.
2. Record the product, mode, locale, device, account state, and timestamp for every answer.
3. Mark whether the brand was absent, named without a visible source, named with a visible citation, or replaced by a competitor.
4. Keep the same scoring rules between runs. If you change prompts, products, or markets, start a new baseline instead of comparing raw percentages.

The simplest formula is:

share of model = answers that visibly name your brand / total evaluated answers under the same test design

Use this only as an observation rule. Do not pretend it is a search-engine-native KPI.

Keep related outputs separate

Do not collapse these into one number:

• brand mention frequency;
• citation frequency;
• referral sessions;
• assisted or last-click conversions;
• revenue or lead quality.

One answer can mention a brand without citing it. One citation can exist without a click. One click can happen without a visible citation in your saved test. If you merge those events, the metric stops being auditable.

Validation workflow

1. Freeze the prompt set and store it with version notes.
2. Re-run under recorded conditions and save the full answer, not just a winning screenshot.
3. Review whether mentions are plain name-drops or are paired with visible source links; move citation QA to AI citations.
4. Check whether the landing pages are technically eligible before blaming the metric. For sc-domain:fennecseo.app from 2026-07-08 to 2026-08-04, /wiki/share-of-model/ returned no visible page row or query row, while /zh/wiki/share-of-model/ returned 1 impression, 0 clicks, average position 1, and no visible query rows.
5. Verify index status separately. On 2026-08-05, Search Console URL Inspection showed both URLs as submitted and indexed, with matching canonicals and successful mobile fetches.
6. If the page may be weak rather than blocked, improve its evidence, boundaries, and next-step links before considering broader consolidation.

Failure modes

• changing the prompt set each run and then calling the percentages a trend;
• merging mentions, citations, clicks, and conversions into one “AI visibility score”;
• treating one product’s output as market-wide share;
• assuming Google AI feature visibility and ChatGPT visibility are measured the same way;
• creating many near-duplicate pages just to chase wording variants or “mentions.”

Next step

Use this page when the job is to compare how often a brand appears across a controlled answer set. If the next question is attribution quality, continue with AI citations. If the next question is frequency of cited answers, continue with Citation rate. If the next question is traffic or outcomes, continue with AI referral traffic. If the page itself may have technical issues, continue with the SEO audit workflow.

Related wiki terms

• AI brand mentions
• AI citations
• Citation rate
• AI referral traffic
• AI crawlers

References

• Google Search Central: AI features and your website
• Google Search Central: Guide to optimizing for Google’s generative AI features
• OpenAI: Overview of OpenAI crawlers