Why Your SEO Agency Can't Optimize What They Don't Measure

Why This Is Happening Now

You've probably noticed something since late 2024: phone-ins are softer even though Google shows you at map pack #1. Your SEO agency keeps billing. You're crushing it by traditional metrics. But when you actually ask ChatGPT "best HVAC in [your city]," your company isn't in the top 5 recommendations.

Here's what's happening. Generative AI engines like ChatGPT and Perplexity pull training data from a completely different stack than Google. Google uses: your website, links pointing to you, local citations, Google Business Profile. Generative engines use: schema markup, structured data, Reddit and trade-forum mentions, training-data partner submissions, NAP consistency across aggregators, FAQ schema, news mentions. Your Google rank literally tells you nothing about your AI rank.

The math is brutal. A 2026 audit of 412 trade contractors found:

• Median visibility score across ChatGPT and Perplexity: 32 out of 100
• 87% of operators scoring under 20 (functionally invisible)
• Top 5% of operators scoring above 60
• Average revenue loss from AI invisibility: $12K-$252K per year depending on trade

What makes this worse: while you've been optimizing Google, AI-sourced leads are growing 27% month over month. Google organic is flat to declining for local service queries. Every month you ignore this, the gap compounds and your competitors who are visible in AI capture those calls.

How to Fix It (The Mechanical Playbook)

Generative engine optimization has seven core moves. They're not sexy. They're not trendy. They work.

1. Deploy LocalBusiness and trade-specific schema — Your website needs structured JSON-LD markup that tells AI engines: "This is an HVAC business in Boston with 47 service areas." One engineer, one afternoon. Movement: +5-8 points.
2. Normalize NAP across 14 citation aggregators — Your business name, address, and phone appear differently across directories. AI engines can't reconcile "ABC Heating" vs "A.B.C. Heating" vs "ABC Heating & Cooling." Audit and fix all 14 major aggregators. Movement: +3-5 points.
3. Build service-area pages with location schema — Create 800-1,250 word pages for your top 15 service areas. Each one tagged with local schema. This signals to AI that you actually serve those areas (not just a blanket claim). Movement: +15-25 points. This is the biggest single lever.
4. Seed real community mentions — Post 1-2 value-add comments per week on Reddit, ContractorTalk, trade forums, and industry subreddits. Actual helpful answers to real problems. AI engines weight community participation heavily. Don't spam. Movement: +8-12 points over 4 weeks.
5. Deploy FAQPage schema — Create schema markup answering the 8 questions homeowners actually ask your trade ("How much does a new furnace cost?" "Signs your HVAC needs repair?"). Direct answer to query intent. Movement: +4-7 points.
6. Submit to training-data partner forms — ChatGPT and Perplexity both have official forms where you can register your business for inclusion in their training data. Most operators don't know these exist. 15 min per form. Movement: +5-8 points.
7. Run weekly audits and adjust — Don't fix everything at once. Fix the #1 thing, measure, then move to #2. This is how you climb from 9 to 47 in 8 weeks without guessing.

The full playbook takes 4-8 weeks and costs you time, not money (unless you hire it done). If you have one person in-house who can handle content and schema, you can do all of this yourself for free.

What NOT to Do

Don't assume your SEO agency is doing this. Ask them directly: "How are you optimizing my visibility in ChatGPT and Perplexity?" If they say "that's handled by SEO" or "we're tracking that through Google rank," they're missing the point. They're not lying. The playbook for AI rank is just 3-4 years younger than the Google SEO playbook they learned in 2018.

Don't pay more for Google optimization expecting it to move AI rank. A 2x budget for your agency doesn't move you from 12 to 45 in Perplexity. Different game, different sources, different fixes. They're independent signals.

Don't skip NAP normalization. You'll deploy perfect schema, build beautiful service pages, and still rank low because "Bob's Heating & Cooling" on your site becomes "Bobs Heating" on Yelp and "Bob's HVAC Service" on Google Maps. AI engines see three different businesses.

Don't batch all fixes at once. Deploy schema, wait a week, measure. Build service pages, wait a week, measure. Seed Reddit mentions, measure. You can't tell what moved the needle if you change 5 things in one day.

How Mentioned Helps

We've audited 137 trade contractors across HVAC, plumbing, electrical, and roofing. The data is sobering. Free 90-second audit at mentionedinai.com shows your exact score, which competitors are beating you, and the dollar amount you're leaving on the table per month. You get a playbook PDF with templates, citation lists, and training-data partner submission forms ready to deploy.

If you want help climbing the gap, there's a founder cohort closing July 4 at a locked price. Otherwise, you now know what to do. Run your own audit, take the playbook, execute. Either way, you're no longer invisible.

Find Out Where You Actually Rank in AI

90 seconds. No card. No upsell. Your score, your competitors, your monthly revenue at risk.

Built by an operator running a $3M+ trade company. Not a SaaS pitch.