The biggest change is who reads your content first. In 2026, a machine reads it before a customer ever does, and that machine decides whether your expertise gets repeated or skipped. Content succeeds now when an AI system trusts it enough to cite it. Page one used to be the finish line. Today it's barely the starting gate.
I've spent more than 30 years in sales and marketing, and I founded LeadValets in 2018. I've watched directories, mobile, local packs, and the review economy all reshape how small businesses get found. None of them moved the ground under content the way this year has.
What does the data say about clicks?
Ahrefs studied 300,000 keywords and found that when an AI Overview appears, the top-ranking page's clickthrough rate falls about 58% (December 2025 data, published February 2026). Read that as an owner, not a marketer. For every 100 calls a #1 ranking used to produce, Google now answers 58 of those people on its own page.
The second finding bothered me more. In a March 2026 analysis of 863,000 search results, Ahrefs found only about 38% of pages cited in AI Overviews also ranked in Google's top 10 for that query, down from roughly 76% the year before (Ahrefs). The link between "I rank" and "I get mentioned" broke in about twelve months. Plenty of contractors who did everything right in 2023 are invisible in the answer their customer actually reads.
Why does generic content fail now?
Because AI can write it in four seconds, and the engines know it. Google's guide to its generative AI search features, last updated July 2026, separates commodity content from non-commodity content. Its example of the weak kind is "7 Tips for First-Time Homebuyers." Its example of the strong kind is a story about waiving an inspection and looking inside a sewer line. That's a plumbing story, and I don't think that's an accident.
That's the whole game for my clients. "5 Signs You Need a New Water Heater" is commodity. Every answer engine already knows those five signs. What it can't produce is what your lead tech found in a 1970s split-level last Tuesday, what the repair cost, and why he talked the homeowner out of a replacement. That detail is the only part worth citing.
The other change people keep missing: there are two kinds of AI search, and they don't eat the same food. Google's AI Mode and AI Overviews can draw on Google Business Profile data. ChatGPT, Perplexity, and Claude build answers from your website, directories, reviews, forums, and video. An agency that reports one "AI visibility" score is averaging two different problems into one meaningless number.
How has LeadValets adapted?
We made four decisions, and each one cost us something.
First, we narrowed. We used to market across five verticals. We repositioned around plumbing and HVAC, then added electrical in September as the third trade. Nobody writes non-commodity content about everything. When a machine is grading you on expertise, depth wins.
Second, we rebuilt our content system around the questions buyers actually ask. We're deep into a 150-plus article build on plumbing and HVAC buyer-intent questions. Every article opens with a direct answer in the first 100 words, ships as server-rendered HTML so crawlers see the text without running JavaScript, and ties every claim to a real credential or a real source. We also refused to spin up a page for every keyword variation. Google names that pattern in its scaled content abuse policy, so we use query fan-out to decide what one page must cover, never how many pages to build.
Third, we stopped selling SEO, answer engine optimization, and the Business Profile as separate services. That's why I built the Search Authority Flywheel, which you can see mapped on our process page. The website strengthens the profile, the profile earns reviews, the reviews shape the AI answers, and the AI answers send people back to the site. Sell those as three line items and you get three half-results.
Fourth, we changed the scoreboard. Rankings and traffic still matter, but we now track citations, share of voice in AI answers, and booked calls. We test prompts by hand on a schedule, separately for Google's AI and the standalone assistants. And new content gets a full 30 days before anyone calls it a failure, because citation lags publication.
Has it worked?
STL Air Pros, a St. Louis HVAC company, went from about two calls a month to more than 111 in roughly eight months, which is documented in that client's case study. One client doesn't prove a law of nature, and I won't pretend it does. It's the result that told me we were building on the right foundation.
We also disclose AI assistance on everything we publish. I sell authority in a field defined by understanding AI. Hiding the tool would be the least credible thing I could do.
What should an owner do this week?
- Ask ChatGPT and Google's AI Mode who to call for your service in your city, and write down who gets named.
- Add one detail to one page that only your crew could know: a real job, a real price, a mistake you prevented.
- Make your business name, phone, and description match everywhere they appear.
- Ask whoever runs your marketing to report AI citations separately from rankings.
Before you change anything, run the free AEO Grader so you have a baseline of what the answer engines can find about you today.