The short version
A growing share of homeowners now ask an AI assistant to recommend a contractor instead of scrolling Google. The businesses that get named have three things: a complete Google Business Profile, a strong review average, and a website an AI can actually parse.
Roughly a third to a half of consumers now use AI assistants at some point when looking for a local business, and only a tiny fraction of businesses ever get recommended. That gap is the opportunity. Ranking in AI answers is not mystical, it just rewards a different set of signals than the ten blue links do.
Why this matters sooner than you think
When a homeowner searches Google, they see ten results and a map pack. They pick from maybe twenty options. When they ask ChatGPT for the best roofer in their town, they get two or three names. That is the entire shortlist.
Going from a page of options to a list of three is the biggest change to local discovery in a decade. It is winner-take-most. If you are not one of the names, you are not in the conversation at all, and the homeowner never sees a list they could have found you on.
The blunt version: Google gives you a chance to be on page one. AI gives you a chance to be one of three. The upside is much larger and so is the cost of being invisible.
The three engines do not use the same sources
This is the part most advice gets wrong. Each assistant has a different idea of who to trust, so optimising for one does not automatically win the others.
Gemini
Gemini sits inside Google, so it reads the Google index and your Google Business Profile directly. If your profile is complete and your reviews are recent, most of the Gemini work is already done. This is the easiest of the three to win and the one contractors most often neglect.
ChatGPT
ChatGPT leans on licensed place data and its search partners rather than crawling Google. Research on its local answers suggests a large majority of the businesses it names first come from third-party place directories. It also draws heavily on discussion sites like Reddit, where real people recommend real contractors.
Perplexity
Perplexity pulls live from the web and cites its sources visibly, which makes it the most responsive of the three to fresh content and recent mentions. If you publish something genuinely useful, Perplexity will notice faster than the others.
The practical takeaway: a complete Google Business Profile covers Gemini, presence in place directories covers ChatGPT, and a real website with current content covers Perplexity.
Your review average is a threshold, not a nice-to-have
This is the single most actionable finding in the whole area. Observers tracking AI recommendations have found the assistants apply what amounts to a rating floor before they will name a business at all.
- ChatGPT tends to recommend businesses averaging roughly 4.3 stars and above
- Perplexity sits a little lower, around 4.1
- Gemini is the most forgiving, around 3.9
If you are sitting at 4.0 with eleven reviews, you are below the line for the assistant with the most users. That is not a branding problem, it is a filter you fail before the conversation starts.
Volume matters alongside average. A business with a handful of reviews reads as unestablished no matter how good the average is. This is why a working review system is not a vanity project. It is the entry requirement.
Do this first: if your average is under 4.3, fixing that beats every other item on this page. Nothing else you do will matter if you are being filtered out before the model even considers you.
Make your site readable to a machine
AI assistants do not look at your design. They parse text and structured data. A beautiful site that says nothing specific is worth less to them than a plain one that states facts clearly.
Structured data
Structured data is a block of code on your site that states, in a format machines read without guessing, what your business is called, where it works, what it sells and who runs it. Every page we build carries it. Most contractor sites carry none, which means the assistant has to infer everything from prose and often gets it wrong.
Answer questions in plain language
Assistants are built to answer questions, so they favour content shaped like answers. A heading that asks what a roof replacement costs in your area, followed by a direct answer in the first sentence, is far more quotable than three paragraphs of buildup.
Be specific and be checkable
Name your towns. Name your trades. State your price ranges. Vague copy that could describe any contractor anywhere gives a model nothing to attach to your name, so it attaches nothing.
Say exactly the same thing everywhere
Models build confidence by corroboration. When your business name, address format, phone number and service description match across your site, your Google profile, your directory listings and your social profiles, the model treats them as one confident entity.
When they disagree, even slightly, confidence drops. A different phone number on your Facebook page than your website is not a small inconsistency to a machine, it is evidence that it might be looking at two different businesses.
Pick one exact format for your business name and use it everywhere, character for character. Same for the phone number and the address.
Get mentioned somewhere that is not your own website
Everything on your own site is a claim. Everything on someone else's site is evidence. Assistants weight the second far more heavily, which is why a business with no external footprint struggles to get recommended regardless of how good its website is.
The realistic list for a contractor: your Google Business Profile, industry directories, your local chamber of commerce, supplier and manufacturer partner pages, local news coverage, and genuine participation in local community forums.
Reddit deserves a specific mention because ChatGPT leans on it noticeably. You cannot spam your way in and attempting it will backfire. But answering questions honestly in your local subreddit, as yourself, is legitimate and it compounds.
How to check whether any of this is working
Do not guess. Audit it directly, and do it roughly once a month.
- Open ChatGPT, Gemini and Perplexity in turn.
- Ask each one the ten questions a homeowner would actually ask. Best roofer in your town. Who to call for an emergency repair. Roughly what a job costs locally.
- Write down whether you are named, in what position, and what it says about you.
- Note which competitors keep appearing. Then go look at what they have that you do not, which is usually reviews.
Do this from a logged-out session where possible, because these tools personalise. Keep the results in a spreadsheet so you can see movement over months rather than relying on your memory of last time.
What this does not replace
AI search is additive. It is not a reason to stop doing the things that already work. Most contractor jobs still come from Google, the map pack, referrals and repeat customers, and that will remain true for a while.
The reason to act now is that the work overlaps almost entirely with work you should be doing anyway. A complete Google profile, a strong review average, a fast readable website and consistent listings serve your regular search rankings just as well. You are not building a separate AI strategy, you are doing local SEO properly and getting a second channel out of it.
Keep reading
The Contractor's Google Business Profile Setup Guide
The profile does more local work than your website.
The Local SEO Checklist for Contractors
The unglamorous list that actually moves rankings.
Frequently asked questions
How do I get my contractor business to show up in ChatGPT?
Start with a complete Google Business Profile and a review average above roughly 4.3 stars, because assistants appear to filter below that. Then add structured data to your website, keep your business name and phone number identical across every listing, and build mentions on sites you do not own.
Is AI search actually sending contractors work yet?
It is a real but still minority channel. Most contractor jobs still come from Google, the map pack and referrals. The reason to act now is that the work required overlaps almost completely with normal local SEO, so you are not spending on a separate strategy.
Do I need different content for AI search than for Google?
No. You need clearer content. Assistants favour direct answers to specific questions, named towns and trades, real numbers, and structured data. All of that helps your Google rankings too.