Back to the visibility tracker

Visibility measurement summary

Measurement date: September 10, 2026, the 48-hour update. 206 completed observations: 197 runs of the frozen v1 prompt set across Perplexity, Google AI Mode, Google organic, ChatGPT, Claude and Gemini, plus nine separate Grok adjunct answers. Every observation carries its own timestamp and visible model or mode label in the run record.

These are observations of Agnostify's own search visibility, self-reported, not client delivery results. Configured locale and country and most underlying model versions were unknown, and Google auto-corrected the branded organic query to agnostic, so no direction against earlier sittings is a controlled comparison.

Latest observation for each measured cell

Each cell's latest observation is dated September 4 or September 10, 2026. Google AI Mode and Gemini ran P01 to P09 only on September 10, so their later prompts keep their September 4 observation.

These figures cover the frozen v1 surfaces only and exclude the nine Grok adjunct answers. The tracker's own headline includes Grok and therefore reads higher for the same date: 45 hits over 239 runs (18.8%) for AI visibility and 45 over 67 (67.2%) for branded identification. Both cohorts come from the same records; only their scope differs.

What the sitting showed

Perplexity returned current prices on the branded price prompt in all three runs, against one of three on September 4. Every commercial discovery run missed Agnostify: 128 on the frozen AI surfaces, plus five of the nine Grok answers. ChatGPT and Gemini again stated a $5,000 minimum that the services page does not carry, and one ChatGPT run resolved the name to a different company. Claude repeated absence-of-records claims. Google's branded organic result was absent from a first page that had auto-corrected the query; that is a miss under an unmatched condition, not a measured rank change.

Coverage and limitations

This cycle assigned 197 v1 runs plus 9 Grok answers, and all 206 were completed. The 235 planned runs in the machine-readable file (Google AI Mode 29, Gemini 48) is the tracker's full frozen-cohort capacity, not this cycle's plan: Google AI Mode and Gemini were assigned P01 to P09 only, so their P10 to P25 cells were out of scope, not failed or unfinished, and keep their earlier observation dates. Two earlier attempts (an anonymous Perplexity query and a ChatGPT collection interrupted when the Mac locked) were excluded and are not scored. The Grok adjunct's nine answers and its separate scoring chat sit outside the frozen v1 comparisons; the scoring chat is not a measurement.

Google AI Mode and organic Google ran in one new signed-out browser Incognito window that was reused across runs, with a fresh thread or fresh navigation per run rather than a fresh window per run; that is a deviation from earlier sittings and one reason no direction is comparable.

Google AI Mode exposes only opaque google.com/goto redirect links; the tracker records each exact observed link address, with its destination unknown and never followed. Captures whose address could not be recovered, and organic result breadcrumbs, are recorded as unknown in the run record, never as no sources. Host-plus-path citations that a surface displayed without a scheme carry an inferred https scheme in the tracker and are labelled as inferred, not observed, in the record.

A brand mention, a linked citation and a vendor recommendation are different outcomes. The underlying transcripts are held in the run record; this page and its machine-readable file summarize the tracker's figures.