Measurement

Share of model

SOM · AI share of voice

Definition

Share of model is a brand’s share of generative answers in a fixed prompt set, on a named engine, in a named period. It is share of voice for chat and AI Overviews. The unit might be “named in the answer,” “cited with a URL,” or “recommended first.” Those three are not the same score.

Vendor dashboards invent slightly different formulas. Write yours down. Run the same prompts, same mode, same login state.

How it works

Fifty prompts, monthly, Perplexity and ChatGPT, logged-out. Count (a) brand named, (b) owned URL cited, (c) earned URL about you cited. SOM named = answers-with-name / 50. Do not blend engines.

Worked example

March: ChatGPT names you in 6/50, Perplexity in 18/50. A competitor is 11 and 22. You gained four Perplexity slots after a trade feature. Report the feature next to the SOM delta so nobody thinks a prompt tweak did it.

How it differs

Citation share uses source slots, not answers. Classic SOV uses journalism or search impressions. A single branded screenshot is a demo, not SOM.

Common errors

Averaging ChatGPT and Perplexity. Using staff accounts with memory. Changing the prompt list every week. Treating a refusal as zero instead of “n/a.”

Sources