Ask an AI who to hire in your industry
Open ChatGPT and ask it who you should hire for what your company does, in your region. Ask the way a buyer would: specific, a little impatient, with a deadline.
Look at the two or three names it gives back. Is yours one of them?
“Our work comes from relationships, not search.” It does, until it doesn’t. Think of the project manager who inherits a vendor list, the procurement lead vetting a new supplier, or the general contractor checking who else can fabricate the rails before the bid closes. More and more of them ask an AI before they call anyone, and what they get back isn’t ten links. It’s a short answer with a few names in it.
So the question isn’t whether AI matters to your business. It’s what the AI read before it left you out.
AI doesn’t invent recommendations. It reads them.
There’s a new industry forming around “AI optimization,” sold as a discipline with its own secret rules. Mostly, it isn’t one.
When an AI assistant answers a “who should I hire” question, it usually searches the web first. Then it builds its answer from what it finds: your website, your Google Business Profile, your reviews, your directory listings, and what other sites say about you. Those are the same signals search engines have read for years.
- The signals: your site, your listing, your reviews.
- One engine: a search engine reads and scores them.
- Two outputs: regular search results and AI answers.
Fix the signals, and both outputs improve. There is no separate trick for AI.
One question becomes many searches
Here’s how a real buyer talks to an AI:
“We need custom lighting for our food processing lines. It has to handle washdowns and meet food safety rules. Who builds this?”
Before it answers, the AI quietly breaks that question into ordinary searches. Google describes this publicly: its AI Mode splits a question into subtopics and runs many searches at once on the user’s behalf. For the question above, those searches might look like:
- custom lighting for food processing lines
- washdown rated LED fixtures
- IP69K food grade light fixtures
- NSF certified lighting manufacturer
- shatterproof lighting for food plants
- USDA compliant facility lighting
- food processing lighting company reviews
- custom industrial lighting quote
Then it recommends two or three businesses that showed up in those results.
If you’re not in those searches, you’re not in the answer.
Notice that the buyer never said “IP69K” or “NSF.” The AI added them, because that’s what the question actually requires. Each search is looking for something specific: a rating, a certification, a regulation, a material, reviews, a way to get a quote. If your website says “Quality industrial lighting since 1987” and not much else, you match none of them.
Your website and your Google listing are two different things
This was the most photographed slide of the talk, so it gets its own section.
| Your Google listing | Your website | |
|---|---|---|
| Where it lives | Google Search and Maps | Your domain, under your control |
| What it holds | Hours, photos, reviews, category | What you do, how, and for whom |
| Its job | It’s the first thing a buyer sees | It’s where the buyer decides |
Most companies keep one up to date and forget the other. In our experience, it’s usually the listing that gets forgotten, because nobody owns it. Marketing thinks IT has it. IT thinks the office manager has it. The office manager left in 2021.
The six signals AI reads about you
No jargon. These are the six things a search engine checks before it decides whether to trust you enough to recommend you.
- Reviews. How many you have, what they say, and whether you reply. A two-star review with a thoughtful owner response often does more for you than silence under a five-star one. The reply isn’t for the person who left the review. It’s for the buyer reading it six months later.
- Consistency. Whether your phone, address and hours match everywhere: your website, your Google listing, industry directories, LinkedIn. When Google can’t tell which phone number is right, it recommends a competitor whose details match. This is the easiest problem to fix and the one most often left broken.
- Speed. How long your site takes to open on a phone, on cell data. Your buyer is often on a job site, not at a desk.
- Labels. Whether the machine knows which fact is which. To a person, “(303) 555-0142” is obviously a phone number. To a machine, it’s a line of text unless your site labels it. Same for your hours, your address, your services and your certifications. Labeled pages tell search engines exactly what each fact is.
- Content. Whether you explain each service or just name it. A page listing eleven capabilities in one sentence looks, to a search engine, like a company that offers maybe two. Each real service deserves a few real paragraphs: what it is, what you work with, who it’s for.
- Mentions. Who else talks about you online: trade associations, supplier pages, local news, project announcements from your customers.
A note for your IT lead: “labels” means structured data, the schema.org vocabulary. It’s usually added as a JSON-LD block in the page head that restates facts already on the page. Nothing visible changes. Google reads it either way.
What a real audit found
We ran our audit on the website of national industrial light manufacturing comany. Big brand, real marketing budget, a site built by people who do this for a living. Here’s what came back.
1. “One forgotten checkbox can make your website invisible.”
A single WordPress setting (“Discourage search engines”) was telling Google and AI crawlers to stay out of the entire site. It had been left on for months, and estimated search traffic fell 47% before anyone noticed. Settings like this often get turned on for a staging site and never turned off, so check your robots.txt and meta robots tags after every launch or migration.
2. “Your security software may be turning away the AI tools you want to be found by.”
The site’s firewall made every visitor pass a JavaScript check, which ChatGPT, Perplexity and Claude’s crawlers can’t do. As a result, Common Crawl, a public web archive many AI models learn from, had captured no pages from the main site in nearly a year. Ask your host or security provider which crawlers they block, and allow the search and AI crawlers you want.
3. “If AI can’t read your website, it describes you using everyone else’s.”
With the company’s own pages blocked, the only sources left for AI answers were its LinkedIn page, a Bloomberg profile and distributor websites. Its 40 years of product specs and roughly 50 customer case studies went unused. Its smaller blog on a separate platform was crawlable and outranked the main site in the top 10 by more than two to one (81 positions vs 34).
Overall score: 33/ 100.
If a site with a national brand behind it misses these, a fabricator whose website was built in 2017 almost certainly does too.
The information provided in your audit led us to having some serious conversations with our current agency.
– Benjamin B, GM of a national industrial lighting manufacturer
There are two kinds of problem
Every issue an audit turns up falls into one of two buckets. One you maintain. The other you rebuild.
| Decay: you maintain it | Structure: you rebuild it |
|---|---|
| It was fine, and the world moved | It was built for a different era |
| Hours and contact details out of date | Capabilities trapped in PDFs and images |
| Reviews left unanswered | A template that won’t let you add pages |
| Speed that slowly got worse | Mobile that’s broken, not just slow |
| Pages nobody has touched in years | No way to add what’s missing |
The distinction matters because the fixes cost very different amounts. Decay is cheap to fix and expensive to ignore. Structure is the reason “let’s just update the site” turns into a six-month project.
It also explains why so many companies redo their website every few years and end up in the same place. A site isn’t finished at launch. It’s operated. Left alone, it slides a little every month as competitors publish, details drift and search engines change what they reward. A rebuild resets the clock. Only ongoing operation stops it.
Five things you can do this week, without us
Everything here is free, and none of it takes more than an hour.
- Reply to every review from the last six months. Especially the bad ones. Keep it short, specific and human.
- Make your phone, address and hours match everywhere. Check your website, Google listing, LinkedIn page and any industry directories you’re in.
- Write a real paragraph for each of your top three services. Say what it is, what you work with, what certifications apply, and who it’s for.
- Open your site on your phone, on cell data, and count. If you get to five before you can read anything, so did your buyer, and they probably left.
- Ask ChatGPT the question your best customer would ask. Then ask it about your company by name, and see what it gets wrong.
The free audit, and why there’s a call
“Free audit” usually means a sales pitch with a PDF attached. That’s a fair suspicion, so here’s exactly how ours works.
We run 25 checks on your website, your Google listing and how you show up across the web. You get a report of about 30 pages. Every finding is ranked by priority, every fix comes with a way to check it worked, and it’s written for the person who runs the company, not the developer.
Then we walk through it with you on a call. That’s the part people ask us to skip, and it’s the part we won’t. A 30-page report with no context gets forwarded, filed and forgotten. On the call, we tell you which findings matter for your business, which ones are noise, and whether you have a decay problem or a structure problem.
Some companies will get a report that says, “You don’t need anything.” We’d rather tell you that than sell you something.