Why it matters
Across a 50-prompt set on industrial lubricants, a distributor appears in 6 answers. That is a 12% AI impression share. Now the distributor has a baseline. Publish a hydraulic-fluid cross-reference table, re-run the same 50 prompts, and watch whether that number moves. Impression share answers a blunt question. In how many of the answers your buyers see do you show up at all? Rankings cannot tell you that, because the AI answer is a single synthesized passage, not a list of ten blue links.
AI impression share vs AI share of voice
They count different things. Impression share is presence: the fraction of answers where you appear, full stop. Share of voice weighs that against competitors, so it is your slice of all brand mentions in the same answer set. You can hold a high impression share and a low share of voice when a rival is named in every answer you appear in.
In practice
An MRO distributor fixes a stable prompt set, say 50 questions covering NLGI grease grades, seal cross-references, and 'best supplier for X' lookups. They run it monthly across ChatGPT, Perplexity, and Google AI answers, then log the share each month. When a new PIM-fed spec page pushes impression share from 12% to 19% on the lubricant prompts, that is the signal the content reached the engines. Treat any single run as illustrative, since answers vary per sample.