Discovery Science · emerging evidence
Are You in the Answer When a Homeowner Asks AI Who to Hire?
The research stage of a remodel has moved upstream into synthesized answers. A homeowner opens ChatGPT or taps the AI Overview above Google and asks for a kitchen cost range or who is good near them, and a short written answer names a couple of firms before any website is opened. This is separate from Google ranking: a remodeler can hold page one for its main term and still be entirely absent from the AI answer written above it, because generated answers are assembled from entity clarity, third-party corroboration and extractable content, not classic ranking signals. Contractor-side adoption of the work that earns those citations is early and uneven, which is the opening. The firm that engineers those signals first is the one the answer can describe with confidence.
Why an AI citation is not the same as a ranking
It is tempting to assume that ranking well means showing up everywhere, but the two outcomes run on different machinery. Classic search returns a ranked list and lets the buyer choose. A generative engine reads the sources, forms a judgment, and returns a verdict with a few names inside it. Being absent from that verdict is not the same as ranking eleventh. It is being left out of the conversation the homeowner is actually having.
The mechanism is structural, not mysterious. Generated answers are stitched together from what the web says about a business: presence across directories and review sites, mentions on pages it does not own, and whether it resolves to one clear entity an engine can trust. A service-area remodeler with a fractured identity across Google, Houzz, Angi and the BBB, thin third-party mentions, and no extractable content on its site gives the engine nothing confident to cite, so it names the competitor it can describe with more certainty.
What the evidence does and does not support
This topic attracts inflated numbers, and two claims are worth separating.
What is documented
The peer-reviewed GEO study (KDD 2024) measured which content levers change whether a source is cited inside a generated answer, and found that adding cited statistics and direct quotations lifted a source's visibility by roughly 30 to 40% on average across the systems tested. Separately, large-sample 2026 analyses report that the majority of AI citations point to sources other than the brand's own site, and that the set of cited domains churns 40 to 60% month over month. Read together: inclusion has identifiable levers, third-party corroboration matters more than self-published claims, and AI visibility is a position to hold rather than win once.
What is not settled
A widely repeated figure says roughly 45% of consumers now use ChatGPT, Gemini or Perplexity for local-business recommendations. It circulates in marketing blogs without a clean primary methodology, so we treat it as directional, not fact, and do not build a case on it. A separate vendor claim that about 87% of independent contractors carry near-zero AI citation share is real reporting, but it was measured on HVAC and plumbing (break/fix), not remodeling, so it does not transfer cleanly to this vertical. In short, the direction is clear and the levers are documented, but the exact remodeling-specific share is something to measure, not assert.
Why remodelers are especially exposed, and especially well-placed
A remodeler is one of the hardest entities for an engine to resolve. Crews travel across many towns with no single storefront address, run several trades under one name, and often list a home office or nothing at all. That ambiguity is exactly what makes an engine hedge. But the same fragmentation that hurts is what creates the opening: almost no independent remodeler has done entity and answer-engine work, so the bar to become the firm an answer can confidently name is low relative to nearly every other channel.
The buyer behavior compounds the opportunity. Cost-research terms carry real volume here, "bathroom remodel cost" at about 33,100 US searches a month and "kitchen remodel cost" at about 22,200, and that top-of-funnel research is precisely where AI answers intercept the homeowner early. A firm that publishes genuine, sourced cost-guide content and resolves to one clear entity is giving the engine something to quote at the exact stage the shortlist is being formed.
How to earn the citation honestly
Resolve to one unambiguous entity across the sources engines read, with correct schema and sameAs links tying verified profiles into one graph. Restructure priority pages to answer the buyer's real question in the opening lines, with clear headings and specific, sourced facts an engine can extract cleanly. Strengthen the third-party footprint across the directories and review platforms that carry weight for remodeling; fabricated mentions are exactly the noise that gets a firm distrusted and skipped.
And then measure it, because no engine publishes this data. A structured read samples a frozen panel of your real buyer questions across each engine repeatedly and records how often you are named, as a rate with a confidence band, stamped with engine, locale and date. Anything less than that is a single lucky check that tells you nothing. That measured share of answer is the starting line, and the number you hold yourself to as the work compounds.
The evidence
Key findings, with their sources
-
Adding cited statistics and direct quotations lifted a source's visibility in generated answers by roughly 30 to 40% on average across the systems tested.
established Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024, arXiv:2311.09735 (peer-reviewed).
-
The majority of AI citations point to sources other than the brand's own site, and the set of cited domains churns 40 to 60% month over month.
emerging Yext / SISTRIX-class AI-citation drift analyses, 2026.
-
Cost-research terms carry real volume: "bathroom remodel cost" draws about 33,100 US searches a month and "kitchen remodel cost" about 22,200.
established Google Ads search volume, US, July 2026 (primary data).
-
A vendor estimate that about 87% of independent contractors carry near-zero AI citation share was measured on HVAC and plumbing, not remodeling, and does not transfer cleanly.
contested DemandConvert, 2026 (vendor, break/fix scope). Cited here as directional only.
Reference
Glossary
- Generative Engine Optimization (GEO)
- The practice of engineering the signals that make a source more likely to be cited inside a generated answer, distinct from ranking a page in a list of links.
- How often a business is named across a panel of buyer questions run repeatedly on each AI engine, reported as a rate with a confidence band and stamped with engine, locale and date.
- Entity resolution
- Whether an engine can confidently identify who a business is, what it does, and where, from consistent signals across the web. Hard for a traveling, multi-trade remodeler.
- Third-party corroboration
- Mentions and citations on pages a business does not own. Large-sample analyses find these carry more weight in AI answers than self-published claims.
Straight answers
Frequently asked questions
Do homeowners really use ChatGPT to find a contractor?
The direction is clear even where the exact number is not. Industry reporting describes homeowners front-loading their research, asking AI assistants for cost ranges and who is good near them before opening a link. A widely repeated figure of roughly 45% using AI for local recommendations circulates without a clean primary methodology, so we treat it as directional rather than fact. What is documented is that cost-research questions carry large volume and that AI answers now intercept them early.
We rank on page one. Why would we be missing from AI answers?
Ranking and being cited in a generated answer are related but separate outcomes. A page-one position can be held while a firm is entirely absent from the AI answer written above it, because engines assemble answers from entity clarity, third-party corroboration and extractable content rather than ranking signals alone. For a traveling, multi-trade remodeler, a fractured entity across directories is a common reason an engine cannot confidently name you.
Can you guarantee ChatGPT or Google AI Overviews will cite us?
No. AI-answer selection is undocumented, volatile and personalized, and engines change how they source month to month. What we commit to is engineering every signal that legitimately moves inclusion, entity clarity, extractable content and third-party presence, and reporting share of answer as measured, including where it stays flat.
How do you measure whether we show up in AI answers?
We freeze a panel of your real buyer questions and run it repeatedly across ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews, then report how often you are named as a rate with a confidence band, stamped with engine, locale and date. Engines are not deterministic, so a single check tells you nothing reliable. That measured share of answer is the starting point for any honest plan.
Provenance
Sources
- Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024, arXiv:2311.09735 (peer-reviewed, established)arxiv.org
- Yext / SISTRIX-class AI-citation drift analyses, 2026 (emerging, large-sample)
- improveit360 and urdesignmag, contractor-selection reporting, 2026 (emerging)improveit360.com
- DemandConvert, independent contractor AI-visibility estimate, 2026 (contested, vendor, break/fix scope)
- Google Ads search volume, US, July 2026 (established, primary data)
Every figure above is attributed to a real, dated source and tagged with its evidence tier. Where a claim could not be verified to a primary source, it is not stated as fact.