AI Visibility

What the change means for SEO, GEO and the measurement of brand presence in generative answers — and why the discipline has no settled owner.

This page reorders the material of the other two sections for SEO / GEO Specialist and AI Visibility Analyst. It collects what used to sit on two separate role pages, because the corpus shows they are one discipline: getting the brand into generative answers and measuring whether it is there.

What changes for AI visibility

  • Disconnection of ranking and citation. Only about 38% of citations in AI Overviews come from the organic top-10, against 76% before; when Google changed the default model in January 2026, about 42% of cited domains were replaced. Being first and being cited are now two distinct goals. See Intent-based interaction.
  • Position is not measurable. Repeating the same question a hundred times, the probability that ChatGPT or Google return the same list twice is below 1 in 100; the same order, below 1 in 1,000. Any tool selling a “ranking position inside AI” is selling an artifact. What is measurable is share of presence — the frequency with which a brand appears across many prompts run many times.
  • The prompt set. People do not write similar prompts even at equal intent (semantic similarity 0.081 across 142 formulations). Build the set by intent, broad, versioned, stable over time, replicated across platforms. Repeat each prompt at least 60–100 times before trusting a data point.
  • Two volatilities. Synchronic (same question, same moment, different answer) is governed by repetition; diachronic (40–60% of cited domains change within a month, per Profound’s longitudinal analysis) is governed by time series. A report that confuses them sums noise and signal.
  • Vendor caution. The AI-tracking market is consolidating fast; before signing, ask how many times each prompt is run, how the set is built, whether methodology is published verifiably, and whether they sell rankings. Search Console has offered an official source since 3 June 2026, but limited to Google surfaces and impressions only.

Responsibility without an owner. This discipline is anything but marginal in the corpus, yet it rarely exists as a role. It measures the output the model returns; Brand & PR guards the input that feeds it; Operations & Data measures downstream traffic. The decision to stay in AI answers therefore has no possible inquest — not for lack of an inquisitor, but because the function that would run it does not exist in most organisations.

What you cannot decide alone