Skip to content
Back to case studies
HVAC · Austin, TX

“Reliable HVAC for a heat pump install”

Mentioned occasionally by Grok, never by ChatGPT or Claude. Two national chains dominated the answers for a market this company knew better.

What we found

The owner had spent years building a reputation for heat pump installs as the category shifted fast. Grok mentioned the business inconsistently; ChatGPT and Claude cited two national chains for install queries, even when the query was clearly regional. The site was thin on install specifics, and nothing on it structured the literal follow up questions buyers were asking.

What we did

  1. 01

    Published a detailed heat pump install guide covering sizing, rebates, ducted versus ductless, and typical install timelines for the region.

  2. 02

    Added FAQPage schema answering the literal queries we observed users asking across platforms.

  3. 03

    Placed two category specific mentions on regional Texas home and trade publications.

  4. 04

    Reconciled service area and install specialty signals across GBP, Bing Places, and six industry directories.

  5. 05

    Added Product schema for the specific heat pump brands the team installs, with warranty and service coverage declared.

What changed

  • ChatGPT began citing the install guide directly for install related follow up questions.

  • Claude picked up the business name within six weeks for regional install queries.

  • Grok mentions stabilized across all install variants.

  • The install guide became the top organic entry page and the second most cited source in attributed AI traffic.

1 → 4 of 5
Platforms citing
0 → 5 of 7
Install query coverage
6
Weeks to Claude pickup
+2
Regional mentions placed

Week by week

  1. Week 1

    Baseline and blueprint

    Twenty install and heat pump queries tested. National chain dominance mapped. Regional intent gap documented.

  2. Weeks 2 to 4

    Foundations live

    Install guide published. FAQPage and Product schema deployed. llms.txt declared install specialties and service area.

  3. Weeks 5 to 8

    Citations and mentions

    Two Texas publication mentions placed. Six industry directories reconciled. ChatGPT began quoting the install guide directly.

  4. Weeks 9 to 12

    Compounding visibility

    Claude picked up the business name for regional install queries. Monthly review shifted next quarter to rebate and efficiency content.

National chains were getting the call because AI kept repeating their names. We had the expertise but nothing structured for a model to actually cite.

Owner, two location HVAC company, Texas.

Questions about this case study

  • What content was published for the HVAC business?
    A detailed heat pump install guide covering sizing, rebates, ducted versus ductless, and typical install timelines for the Austin region. The guide became the top organic entry page and the second most cited source in attributed AI traffic.
  • Which schemas were deployed?
    HVACBusiness (via LocalBusiness), Service per specialty, FAQPage on the install guide, and Product schema for the specific heat pump brands the team installs with warranty and service coverage declared.
  • How did Claude pickup timing compare to ChatGPT?
    ChatGPT began citing the install guide within four weeks. Claude took six weeks to pick up the business name for regional install queries.
  • Which directories were reconciled?
    GBP, Bing Places, and six industry directories relevant to HVAC — service area and install specialty signals were reconciled to a single source of truth.
  • Did national chains stop dominating the category?
    No. National chains continued to appear for generic install queries. The business displaced them on regional and specialty queries where the install guide and schema carried the signal.

Representative case study. Industry, location, and specifics are illustrative composites drawn from recurring patterns in our work. AI answer engines are probabilistic; actual results vary by category, competition, and baseline.

See where your business actually stands.

Thirty minutes, real queries, real findings.