Wikimotive Research • 2026
AI Visibility Study 2026
WHAT WE TESTED
Metro × Rural
Market context comparison • ChatGPT results • March–September 2026
1,431 source citations + 143 dealer recommendations • Cumulative totals across all recorded sampling rounds
01
THE QUESTION CHANGES THE SOURCE
Owned-site share of citations, by prompt
● Metro
● Rural
“Tell me about…”
“Are they good?”
“Reviews”
Evaluative wording shifts attention to third-party voices.
Rural complaint prompts: 24/24 citations were review pages.
Metro group-brand query: 75% owned citations.
02
REPUTATION IS A COMMON THREAD
recorded Google ratings met 4.2+ among recommended dealers
Google ranges: Metro 4.2–4.7 • Rural 4.2–4.8
Ratings available for 140 of 143 recommendations. No minimum threshold was established.
03
CONTENT & INVENTORY DO DIFFERENT JOBS
had written content
157 of 182 Metro dealer-page citations
had on-page inventory
126 of 182 Metro dealer-page citations
ACROSS 185 ASSESSABLE DEALER-PAGE CITATIONS
■ Both: 101
■ Inventory Only: 27
Content-only pages surfaced, too.
57 had text without stock; none had neither.
Pooled sample: 182 Metro + 3 Rural observations; these are associations, not isolated effects.
04
SHOPPING LISTS DOMINATE INVENTORY SOURCES
Share of inventory citations going to search-results pages
Metro — 79% · 769/970
Rural — 98% · 253/257
AUTOTRADER + CARS.COM: SHARE OF INVENTORY CITATIONS
Metro
Rural
Matched comparison: same 22 prompts, July 22 + August 12. Citation bases: 126 M / 130 R.
Metro marketplace share grew over time
March–May → July–August • same 28 inventory prompts
Same prompt set, different market. Same market, different date.
A one-time snapshot does not describe a stable AI leaderboard.
05
WHERE THE ANSWER BREAKS
Wrong intent. Wrong business. More friction.
“Colorado” → real estate
9 irrelevant citations
“Equinox” → a gym
3 irrelevant citations
Rural → another dealer
3 complaint-source citations
Possible implication: shopping frustration → a return to traditional search.
Interpretation, not measured consumer behavior. These studies did not measure switching or conversion.
THE DEBUNK BRANCH
NO UNIVERSAL SHORTCUT
Requirements & advantages that did not hold
LLMS.txt
No observed benefit.
Google Confirmed May 20262
Special schema
No observed ranking advantage.
No consistently winning schema type.1
Site platform
No consistent winner.
Provider did not explain a stable order.1
FAQS / MAPS REQUIRED?
had no FAQs
180 / 185
had no map / directions
161 / 185
Cited dealer pages; absence did not prevent citation.
BIGGEST STOCK WINS?
First had the most inventory.
15 of 43 complete three-dealer result sets.
GREAT YELP REQUIRED?
Yelp range for Metro recommendations
with ratings recorded (n=113).
TWO SITES, DIVIDED ATTENTION
⇄
Second site
An OEM-approved site competes with a non-compliant second site, dividing visibility and maintenance attention.
LOCAL WINNERS BEHAVE DIFFERENTLY
different first-place dealers
across 13 Metro rounds
Rural tests: same dealer first
One recurring competitor led every Rural test.
A DEALER NAMED ≠ A DEALER SITE LINKED
The answer can name dealers while citing marketplaces.