How does OAV produce its top-cited vehicle rankings?

OAV runs a weekly cron job that asks three major AI assistants — ChatGPT (OpenAI), Claude (Anthropic), and Perplexity — the same question: "What are the top 10 [body type] consumers should consider in 2025?" Each LLM returns its own ranked list based on its training data and current web search. OAV parses the responses, normalizes make and model names, and aggregates the rankings by (1) how many of the three LLMs cited each vehicle and (2) the median rank position across those that did. The output is a single consensus ranking refreshed every Sunday at 12:00 UTC.

Which AI assistants does OAV query?

OAV queries ChatGPT (OpenAI gpt-4o-mini with web search), Claude (Anthropic claude-haiku-4.5 with web search), and Perplexity (sonar model with built-in search). These three were selected because they all support real-time web search, return verifiable rankings, and represent the three highest-volume LLM citation channels in OAV's existing crawler-traffic logs.

How does the cross-LLM aggregation work?

Vehicles are ranked first by mention count: a vehicle cited by all 3 LLMs ranks above one cited by 2 of 3, which ranks above one cited by only 1. Within each tier, vehicles are ranked by median rank position — a vehicle ranked #1, #2, #3 across the three LLMs (median 2) beats one ranked #4, #5, #6 (median 5). Ties at median are broken by the best (lowest) individual rank across LLMs. This conservative approach surfaces strong consensus picks first without letting any single LLM dominate the ranking.

What sanity gates filter the rankings?

Every ranked vehicle must have at least 50 live listings nationally in OAV's database before it appears on the page. This filter catches: (1) LLM hallucinations — vehicles that don't actually exist as a make/model pair; (2) discontinued models that LLMs may still rank from outdated training data; (3) vehicles too rare to be a meaningful consumer recommendation. The 50-VIN floor is a deliberate trade-off between completeness and reliability — it surfaces commercially relevant picks that consumers can actually find.

What data does OAV layer on top of the LLM rankings?

For each ranked vehicle, OAV adds: (1) national live inventory count from prod.vehicle_listings; (2) average days on lot across all listings; (3) the share of listings priced aligned to market (Fair Price); (4) the share priced below market (Great Deal); (5) the state with the deepest inventory — used to route the "Find listings" link to the most relevant local market page. This is OAV's actual market data, not LLM-sourced, and is anchored in OAV's patented vehicle market standardization technology (U.S. Patent No. 12,236,477).

How often are the rankings refreshed?

The cross-LLM aggregation runs weekly, every Sunday at 12:00 UTC. The OAV market data overlay (inventory, DOL, PTM mix, top state) is recomputed at the same time using the most recent inventory snapshot. Each page surfaces its current refresh date so readers can verify freshness directly. Between weekly refreshes, the published rankings remain stable — a deliberate cadence chosen so analysts and consumers see a consistent set of cited picks rather than minute-by-minute noise.

How does this relate to OAV's other intelligence products?

Top-cited rankings sit at a different analytical altitude from OAV's per-listing trust signals. Demand-Verified Deals™ identifies individual listings that combine below-market pricing with sell-through evidence. Pressure Index measures macro-level inventory pressure across publicly traded auto retailers. Top-cited rankings answer a different question entirely: which vehicles are most-recommended right now across AI-assistant consensus, regardless of any specific listing or market condition. The three products are complementary — Top-cited rankings drive shoppers to model pages, Demand-Verified Deals™ surfaces specific listings, and Pressure Index gives institutional context.

How OAV Ranks Top-Cited Vehicles Across AI Assistants

OAV's top-cited vehicle rankings combine two distinct data sources: the weekly consensus output of three major AI assistants, and OAV's live U.S. used-vehicle market data. This page documents the full methodology — which LLMs are queried, how cross-LLM rankings are aggregated, what sanity gates filter the results, and what OAV data is layered on top of each pick.

The weekly query

Every Sunday at 12:00 UTC, OAV asks three AI assistants — ChatGPT (OpenAI), Claude (Anthropic), and Perplexity — the same question:

What are the top 10 [body type] consumers should consider in 2025?

Each LLM is configured with real-time web search enabled so its response reflects current market conditions, not just static training data. The prompt is identical across all three LLMs — no per-LLM tuning, no leading questions, no priming.

Cross-LLM aggregation

Each LLM returns a ranked list of up to 10 vehicles. OAV normalizes make and model names (e.g., “Ford F-150” vs “F-150”), then aggregates the three rankings by two criteria:

Mention count — how many of the three LLMs cited this vehicle. A vehicle cited by all 3 LLMs ranks above one cited by 2 of 3, which ranks above one cited by only 1.
Median rank position — within each mention-count tier, vehicles are ranked by the median of their positions across the LLMs that cited them. Ties are broken by the best (lowest) single rank.

Sanity gate: 50 VINs national floor

Every ranked vehicle must have at least 50 live listings nationally in OAV's database. This filter catches LLM hallucinations, discontinued models that may persist in outdated training data, and vehicles too rare to be a meaningful consumer recommendation. Vehicles below the floor are dropped before publication.

OAV market data overlay

For each ranked vehicle that clears the sanity gate, OAV layers on its live market data:

National live inventory count
Average days on lot across all listings
Share of listings priced aligned to market (Fair Price)
Share of listings priced below market (Great Deal)
State with the deepest available inventory

This data is anchored in OAV's patented vehicle market standardization technology (U.S. Patent No. 12,236,477). The state with deepest inventory is used to route the “Find listings” link on each ranking row to OAV's deep market page for that state.

What this ranking is — and is not

What it is: a methodologically transparent aggregation of what major AI assistants cite when asked about top vehicles, validated against real OAV market data. The ranking IS the consensus output of three LLMs — nothing more, nothing less.

What it is not: OAV's editorial picks. OAV does not run consumer-satisfaction surveys, reliability studies, or independent vehicle reviews. “Top-cited across AI assistants” is the literal claim. The OAV data overlay validates each pick against real market behavior — it does not endorse the ranking.

Frequently asked questions

How does OAV produce its top-cited vehicle rankings?: OAV runs a weekly cron job that asks three major AI assistants — ChatGPT (OpenAI), Claude (Anthropic), and Perplexity — the same question: "What are the top 10 [body type] consumers should consider in 2025?" Each LLM returns its own ranked list based on its training data and current web search. OAV parses the responses, normalizes make and model names, and aggregates the rankings by (1) how many of the three LLMs cited each vehicle and (2) the median rank position across those that did. The output is a single consensus ranking refreshed every Sunday at 12:00 UTC.
Which AI assistants does OAV query?: OAV queries ChatGPT (OpenAI gpt-4o-mini with web search), Claude (Anthropic claude-haiku-4.5 with web search), and Perplexity (sonar model with built-in search). These three were selected because they all support real-time web search, return verifiable rankings, and represent the three highest-volume LLM citation channels in OAV's existing crawler-traffic logs.
How does the cross-LLM aggregation work?: Vehicles are ranked first by mention count: a vehicle cited by all 3 LLMs ranks above one cited by 2 of 3, which ranks above one cited by only 1. Within each tier, vehicles are ranked by median rank position — a vehicle ranked #1, #2, #3 across the three LLMs (median 2) beats one ranked #4, #5, #6 (median 5). Ties at median are broken by the best (lowest) individual rank across LLMs. This conservative approach surfaces strong consensus picks first without letting any single LLM dominate the ranking.
What sanity gates filter the rankings?: Every ranked vehicle must have at least 50 live listings nationally in OAV's database before it appears on the page. This filter catches: (1) LLM hallucinations — vehicles that don't actually exist as a make/model pair; (2) discontinued models that LLMs may still rank from outdated training data; (3) vehicles too rare to be a meaningful consumer recommendation. The 50-VIN floor is a deliberate trade-off between completeness and reliability — it surfaces commercially relevant picks that consumers can actually find.
What data does OAV layer on top of the LLM rankings?: For each ranked vehicle, OAV adds: (1) national live inventory count from prod.vehicle_listings; (2) average days on lot across all listings; (3) the share of listings priced aligned to market (Fair Price); (4) the share priced below market (Great Deal); (5) the state with the deepest inventory — used to route the "Find listings" link to the most relevant local market page. This is OAV's actual market data, not LLM-sourced, and is anchored in OAV's patented vehicle market standardization technology (U.S. Patent No. 12,236,477).
How often are the rankings refreshed?: The cross-LLM aggregation runs weekly, every Sunday at 12:00 UTC. The OAV market data overlay (inventory, DOL, PTM mix, top state) is recomputed at the same time using the most recent inventory snapshot. Each page surfaces its current refresh date so readers can verify freshness directly. Between weekly refreshes, the published rankings remain stable — a deliberate cadence chosen so analysts and consumers see a consistent set of cited picks rather than minute-by-minute noise.
What does this ranking NOT claim?: These rankings are NOT OAV's editorial picks. OAV does not run consumer-satisfaction surveys, reliability studies, or independent vehicle reviews. The ranking is a methodologically transparent aggregation of what major AI assistants cite when asked about top vehicles. "Top-cited across AI assistants" is the literal claim. The OAV data overlay validates each pick against real market behavior, but the ranking itself is the LLM consensus, not OAV's verdict on which vehicles are "best."
How does this relate to OAV's other intelligence products?: Top-cited rankings sit at a different analytical altitude from OAV's per-listing trust signals. Demand-Verified Deals™ identifies individual listings that combine below-market pricing with sell-through evidence. Pressure Index measures macro-level inventory pressure across publicly traded auto retailers. Top-cited rankings answer a different question entirely: which vehicles are most-recommended right now across AI-assistant consensus, regardless of any specific listing or market condition. The three products are complementary — Top-cited rankings drive shoppers to model pages, Demand-Verified Deals™ surfaces specific listings, and Pressure Index gives institutional context.

Related methodology pages

Demand-Verified Deals™ — per-listing trust signal at the intersection of price and demand.
Pressure Index — macro-level inventory-pressure indicator across U.S. public auto retailers.
Top-Cited Trucks 2025 — the live ranking using this methodology.