Measurement
How often AI answer engines cite OAV
As of the run of September 22, 2026 UTC, AI answer engines cited an app.oav.io page in 30.3% of the 800 web-search queries OAV asked them, 242 of 800. The queries are about best-priced vehicles across major U.S. metros and body types. This figure comes from instrument version 4, in effect since September 22, 2026 UTC. Shares are comparable within an instrument version, not across versions.
Share by engine, run of UTC
| Engine | Queries asked | Cited an OAV page | Share |
|---|---|---|---|
| OpenAI | 160 | 75 | 46.9% |
| Claude | 160 | 9 | 5.6% |
| Perplexity | 160 | 134 | 83.8% |
| Gemini (grounded) | 110 | 0 | 0.0% |
| Gemini (ungrounded) | 50 | 0 | 0.0% |
| Grok | 160 | 24 | 15.0% |
| All engines | 800 | 242 | 30.3% |
Engines report sources differently. Some list every source they consulted, others only the sources they cite inline, so a per-engine share is comparable for one engine over time, not between engines. A run counts a query as citing OAV when the engine's answer or its listed sources include an app.oav.io page.
Complete runs on instrument version 4
One complete run has been made on instrument version 4, in effect since September 22, 2026 UTC. A second run gives a trend.
| Run date (UTC) | Queries asked | Cited an OAV page | Share |
|---|---|---|---|
| 800 | 242 | 30.3% |
Queries asked counts every query that returned an answer in that run. Queries that returned an error are excluded from both columns.
Earlier instrument versions
These runs were produced by earlier versions of the instrument, which asked a different set of engines or counted citations differently. Their shares are not comparable with the figures above or with each other across versions. They are listed so the history is checkable, not so it can be trended. The 12 most recent are shown; 2 older runs are not listed.
| Instrument version | Run date (UTC) | Queries asked | Cited an OAV page | Share |
|---|---|---|---|---|
| Version 3 | 640 | 214 | 33.4% | |
| Version 3 | 640 | 204 | 31.9% | |
| Version 2 | 640 | 83 | 13.0% | |
| Version 2 | 639 | 79 | 12.4% | |
| Version 2 | 640 | 60 | 9.4% | |
| Version 2 | 640 | 67 | 10.5% | |
| Version 2 | 640 | 62 | 9.7% | |
| Version 2 | 639 | 63 | 9.9% | |
| Version 2 | 480 | 75 | 15.6% | |
| Version 2 | 480 | 71 | 14.8% | |
| Version 1 | 480 | 70 | 14.6% | |
| Version 1 | 480 | 60 | 12.5% |
Instrument versions
The instrument is versioned. A version changes whenever the set of engines asked changes, or a correction changes what a run counts. Shares are comparable within a version, not across versions. This is the full history, so a move in the headline figure can be read against what moved in the instrument.
- Version 1 · introduced UTC · 3 engines asked
First instrument. Three engines asked: OpenAI, Claude and Perplexity. - Version 2 · introduced UTC · 4 engines asked
Gemini added, asked in grounded and ungrounded modes and reported as two rows. The engine set changed, so shares are not comparable with version 1. - Version 3 · introduced UTC · 4 engines asked
No engine change. A counting defect affecting one engine was corrected: that engine's citations of OAV had been recorded as absent since version 2. Because the correction changed what a run counts, the version was bumped and the first corrected run starts a new epoch. The step up in overall share at this version is the correction, not a change in how often OAV is cited. - Version 4 · introduced UTC · 5 engines asked · current
Grok added. The engine set changed, so shares are not comparable with version 3.
What this page is not
This page is not a ranking of answer engines, not a measure of traffic to OAV, and not a claim about any other website. It reports one thing: how often OAV's own web-search queries to these engines came back with an app.oav.io page among the answer or its listed sources, on a stated date, on a stated instrument version.
Engine names are used to identify the product asked; OAV has no partnership, integration, affiliation, or endorsement relationship with any of them.