Methodology: The 2026 AI Visibility Index for Trade Contractors

Published 2026-06-03 · Mentioned · mentionedinai.com

1. Sample

100+ companies audited across four trades (HVAC, plumbing, electrical, roofing) plus 10 leading field-service software vendors. Geographic coverage: 17 metro markets across 14 US states. Sample composition mirrored 2024 BLS data on trade contractor distribution by metro size.

2. Query design

30 standardized queries per company, structured across three categories:

Best in city (10 variants): "best HVAC in Phoenix", "top-rated HVAC Phoenix", "who is the best HVAC in Phoenix", etc.
Problem-driven (10 variants): "my AC is leaking, who should I call in Phoenix", "best HVAC for emergency repair Phoenix", etc.
Near-me (10 variants): "HVAC near 85001", "HVAC contractors near downtown Phoenix", etc.

City and trade tokens were varied per audit to match the company's actual service area. Queries were chosen to mirror real homeowner question patterns observed in 2025-2026 trade industry surveys.

3. AI engines

Each query was run across three AI search engines:

ChatGPT (GPT-5, with web browsing enabled)
Perplexity (default model, Pro tier)
Gemini (Gemini 2.0 Flash with Search grounding)

Queries were run from US-based IPs across 4 geographic regions to control for personalization. Each query was repeated 3 times per engine per region; the modal answer was retained.

4. Scoring

For each (query, engine) pair:
· 1 point if the company is named anywhere in the response
· 2 points if the company is named first
· 0 points if not named

Maximum raw score: 30 queries x 3 engines x 2 points = 180. Normalized to 0-100 scale.

5. Time-series reconstruction (2024 + 2025)

2024 and 2025 baselines were reconstructed by running the same query set against archived AI engine outputs (where available) and standardized re-runs of the engines as they existed in those periods. Reconstruction error margin: +/- 3 points. 2026 baseline is direct measurement, no reconstruction.

6. Independent review

Independent methodology review is being arranged. Mentioned is recruiting three practitioners with no equity in the company (an AI search practitioner, a structured-data specialist, and a trade industry operator) to spot-check the query set, scoring formula, and raw audit output. Reviewer credits and notes will be published when the review is complete. Trade publications and academic researchers can request the raw audit output at partnerships@mentionedinai.com.

6a. Conflict-of-interest disclosure

Mentioned (the publisher) was audited via the same methodology against the same query set and scored 30.8/100, placing the publisher in the same bracket as the median operator on this index. Per methodology, the publisher is excluded from public leaderboards. Mentioned will re-audit itself quarterly and publish score movement transparently. Next self-audit: 2026-08-28.

7. Limitations

AI engine outputs are non-deterministic. Score has +/- 2 point margin even with 3 repetitions per query.
Personalization (geography, browser history) cannot be fully controlled even with IP rotation.
Reconstruction error margin (+/- 3 points) applies to 2024 + 2025 only.
Sample is biased toward operators with public web presence; small contractors without websites underrepresented.

8. Data availability

Full dataset is downloadable at:

CSV: mentionedinai.com/benchmark.csv
JSON: mentionedinai.com/benchmark.json
License: CC-BY-4.0 with attribution to Mentioned

Citation

Beyer, N. (2026). The 2026 AI Visibility Index for Trade Contractors. Mentioned. https://mentionedinai.com/benchmark