The HVAC Software AI Visibility Report, July 2026

When a contractor asks an AI assistant which HVAC software to buy, the answer is a short list of names assembled from third-party sources. This report measures who is in those answers. We ran the same fixed basket of 40 real buyer questions for 12 HVAC/field-service software vendors, the same way for every vendor, and counted who gets named.

This is measurement, not a quality ranking. A vendor's position here says how often AI answers name it on our fixed question set, and nothing else. No vendor paid to be scanned, named, or excluded, and inclusion is not an endorsement.

The one number

1 of 12 HVAC software vendors we scanned are never named in a single one of its 120 captured answers from Perplexity, this edition's live engine (40 buyer questions x 3 independent passes, verified at pass level in the scan log).

The median vendor holds a 6.25% share of answers. The most-named vendor, ServiceTitan, holds a 50.0% share of answers and is named in 38 of 40 buyer questions. The middle of this market is thin: seven of the twelve vendors hold a share of answers of 6.25% or less. Five of those are named too rarely to earn any share under majority scoring, and one, as above, is never named at all.

League table: share of answers

Share of answers = the percentage of the basket's 8 question clusters where the vendor is named under majority rule (full and half credit per the banding in the methodology below). Ties break on buyer questions named, then alphabetically. Named in buyer questions = the count of the 40 questions where the vendor was named in at least 2 of 3 passes (majority rule) on the live engine.

Position	Vendor	Share of answers	Named in buyer questions (of 40)	Engines naming it
1	ServiceTitan	50.0%	38	Perplexity
2	Housecall Pro	37.5%	32	Perplexity
3	Jobber	31.25%	30	Perplexity
4	FieldEdge	18.75%	24	Perplexity
5	Workiz	18.75%	14	Perplexity
6	Service Fusion	6.25%	18	Perplexity
7	FieldPulse	6.25%	9	Perplexity
8	BuildOps	0.0%	10	Perplexity
9	simPRO	0.0%	5	Perplexity
10	Kickserv	0.0%	3	Perplexity
11	ServiceM8	0.0%	1	Perplexity
12	Commusoft	0.0%	0	none

Per-engine breakdown

Engine	Status	Vendors named at least once	Never named on this engine	Median share on this engine
Perplexity	scanned	11 of 12	1	6.25% (median across all 12 scanned vendors, matching the league table)
ChatGPT	unavailable this study	not scored	not scored	not scored
Google AI Overviews	unavailable this study	not scored	not scored	not scored
Copilot	unavailable this study	not scored	not scored	not scored

Engines marked unavailable were not reachable in a reproducible, signed-out form for this study. They are disclosed as unavailable, never approximated or faked. When an engine comes online it joins a future edition against the same frozen basket.

Methodology

Fixed basket. 40 real buyer questions across 8 intent clusters (category head, small-contractor, competitor-displacement, dispatch/scheduling, QuickBooks integration, mobile app, flat-rate pricing, adjacent trades), frozen verbatim in code. Methodology fingerprint: 6dbb8f757853.
One scan per vendor, same ruler. Every vendor was scored from its own scan of the identical basket. Scan window: 2026-07-17. Region: US / en-US (national).
Passes and majority rule. Every question ran 3 independent passes per engine. A vendor counts as NAMED on a surface only when it is named in the answer prose on a majority of passes; single-pass results are noise.
Cluster banding. A cluster scores 1.0 when the vendor is named on at least 60% of its surfaces, 0.5 between 20% and 59%, and 0 below 20%. Share of answers = sum of cluster scores divided by 8.
Engines. Requested: Perplexity, ChatGPT, Google AI Overviews and Copilot. Covered in this edition: Perplexity. Unavailable and disclosed as such: ChatGPT, Copilot and Google AI Overviews.
Coverage split, stated plainly. Cited's free scan runs an 8-query quick cut of this basket (1 question per cluster). This study and the paid Fix Plan run the full 40-query basket. Quick-scan numbers are never mixed into this report.
Audit trail. Every scan emits a raw per-pass log (every answer, every citation, timestamped), so every number in this report can be reproduced and audited.

Limitations, stated plainly

Only Perplexity was covered in this edition, so the headline number is really share of Perplexity answers. We say that here rather than letting the title imply more.
3 passes per question is a thin sample for non-deterministic engines. It is the floor at which majority rule means anything; more passes would tighten the estimate.
NAMED detection is word-boundary string matching against a tracked vendor list in the answer prose. Vendors outside the tracked universe of a given scan are not detected on that scan.
This is a snapshot. AI citations churn naturally every 4 to 8 weeks, so treat the table as this month's reading on a frozen ruler, not a permanent standing.
The vendor universe is the 12 vendors we scanned, not the whole market. A vendor missing from the table was not scanned, which is different from never named.
Freezing the basket buys comparability at the cost of drift from how buyers phrase things over time. Deliberate trade-off.
This report measures visibility only. It does not test what changes visibility, and it makes no claim about any vendor's product quality.

Published by Cited. We measure whether AI answers name a vendor on a fixed, reproducible basket. We never promise a specific rank, inclusion in any source, or any specific answer outcome; nobody controls what an AI engine outputs.

Methodology fingerprint: basket 6dbb8f757853 . engines Perplexity, ChatGPT, Google AI Overviews and Copilot . 3 pass(es) per surface . scan window 2026-07-17. Future editions re-run the identical basket, because movement only means something measured against a frozen ruler.