Your Site Sells Specs. AI Recommends Someone Else
A buyer types a question into ChatGPT: "Which one of these is right for a business like mine?" The answer comes back assembled from a Reddit thread, a review site, and the distributor who resells your product at a markup. Your name is missing, even though you make the thing. That happens because your website is a spec sheet, and the buyer asked a decision question, and the two are not the same thing.
This piece defends one position for the next 2,000 words: you do not have a keyword problem, a schema problem, or an E-E-A-T problem. You have an answer-coverage problem. The questions that decide a sale already live in your sales calls, your support tickets, your DMs, and your own site-search box. You answer them out loud every week, on repeat. You have never published them. Fix that, and you stop being a catalog that AI skips over.
AI Builds the Recommendation From Everyone Except You
Watch what an AI assistant actually does when someone shops. It does not read your spec table and get inspired. It hunts for the reasoning a buyer needs: will this fit my case, is it covered under my situation, how fast can I get it, how does it stack against the obvious alternative. When your pages only list dimensions, materials, and part numbers, the model goes looking for that reasoning somewhere else. It finds it on forums and third-party roundups, because those places argue the choice while you only describe the product.
Bill Hunt named this gap plainly in Search Engine Journal on July 22, 2026. His framing is the whole problem in one line.
The cost of that gap is not abstract. It shows up as a recommendation you are not in. When a distributor, a review aggregator, or a competitor publishes the "why choose this" reasoning and you publish the "here are the specs" list, the model treats them as the authority on your own product. Hunt put a sharper point on it, and it should sting a little.
I have watched this exact dynamic play out in the reselling problem too. If you have ever wondered why a Reddit thread and a reseller keep out-ranking you on your own offer, this is the mechanism. They wrote the decision. You wrote the datasheet.
You Keep Auditing Keywords When You Should Audit Answers
Open any "how to show up in AI" article and you will get the same four-item recipe: build E-E-A-T, collect reviews, add schema, and answer real questions. All four are correct. All four are also useless the way they are usually delivered, because not one of them tells you which questions are worth publishing or where to get them. "Answer real questions" is where every one of those articles quietly stops. The owner nods, closes the tab, and still has no list.
That missing list is the whole point of this piece. The concrete step nobody hands you is this: harvest the specific pre-purchase decision questions your buyers already ask, from the four places you already own, and publish the answers. Objections. Comparisons. "Will it work for my case." Coverage, fit, speed. Those questions are not hypothetical, and you do not need a research budget to find them, because you are already answering them all day. This post is the deep dive on Step 2 of my broader step-by-step AI visibility checklist, the one that says rank for the real questions customers ask. Read that for the full sequence. Read this for the part where you actually build the list.
Owners get this wrong for a predictable reason. Auditing keywords feels like progress because tools spit out volumes and difficulty scores. Auditing answers feels like admitting you have gaps, so people avoid it. Hunt draws the line between the two directly, and it reframes what "add another blog post" is even for.
Read that twice. The problem is rarely volume. It is that the knowledge you carry in your head has never been written down where a machine, or a buyer at 11pm, can reach it.
The Clicks Are Already Leaving, and Specs Will Not Hold Them
This is not a "someday" shift. The traffic math already moved. In a randomized field experiment run on 1,065 Chrome users with data collected January 7 to February 10, 2026, researchers Agarwal at the Indian School of Business and Sen at Carnegie Mellon's Heinz College found that Google's AI Overviews cut outbound organic clicks by 39.8% and pushed zero-click searches up 34.5%. Translation for your month: four in ten people who would have clicked through now get their answer without ever landing on a site.
One detail from that same study matters more than the headline number. Bounce rate, back-button rate, and time-on-site did not differ. People are not clicking and bouncing in frustration. They are getting a satisfying answer inside the AI box and moving on, content. That is the part that should reset your strategy. You are not competing for a click anymore. You are competing to be the source the answer is built from.
And that window is narrowing on the search side too. Ahrefs reported on February 4, 2026 that click reduction for top-ranking pages had grown to 58%, up from 34.5% the previous April. BrightEdge data compiled by SQ Magazine put AI Overviews on 48% of tracked queries as of early 2026, up from 31% a year earlier. Half your queries now carry an AI answer on top, and the top spot leaks more than half its clicks. Ranking number one is no longer the finish line. It is a coin flip on whether anyone sees you at all.
So who does the AI cite? Not you, mostly. Surfer SEO analyzed 46 million AI Overview citations and found YouTube at 23.3%, Wikipedia at 18.4%, and Google's own properties at 16.4%. Third-party sources dominate the answer. If you have never checked what AI actually says when someone asks about your business, that is the first hour of work, because you cannot fix a recommendation you have never read.
| Product data (what your site has) | Decision data (what the buyer and AI need) |
|---|---|
| Dimensions, materials, part numbers | "Will this hold up for a clinic that runs it 12 hours a day?" |
| Feature list and tiers | "Which tier is right if I only need it for two sites?" |
| Price and SKU | "Is it covered for my use, and how fast can I get it?" |
| "Compare our models" grid | "How do you compare to the alternative everyone mentions?" |
Four Places You Already Own That Hold Your Real Questions
Stop brainstorming questions in a conference room. Your buyers already handed them to you, for free, in four sources you control. The job is harvesting, not inventing. Hunt tells one story that should light a fire under this. A company mined its own internal site-search logs for revenue-related queries and generated $6.8 million. One query had been searched more than 100,000 times, asking how to convert a single-day pass into a multi-day pass. The site's FAQ answered it with a one-word "Yes," which ended the search instead of opening the next step. The demand was screaming in the logs. Nobody was reading the logs.
The harvest goes in this order. Each source costs you nothing but an afternoon of reading.
Once you have the list, publishing is the easy part. Write the answer the way you say it on the phone, in plain sentences, with the caveat included. If a product genuinely does not fit a use case, say so; that "no" earns more trust than a vague "it depends," and it keeps the wrong buyer from a return. A decision answer backed by a real number or a named source is exactly what a model reaches for, which is why verifiable proof beats adjectives every time.
My confidence here comes from a specific place. Across 17 years in search and 300+ businesses, the pattern holds: the businesses that get chosen are the ones whose answers are public before the first phone call. The strongest result of my career, a London ADHD clinic booked solid for three straight months until they hired more specialists and outsourced the overflow, was won in healthcare, a market where every buyer arrives carrying fit, coverage, and timing questions. Specs never book anyone. Answers do. Your buyers make the same call before they ever reach your website, which is exactly why the answers have to be sitting out in public first.
Measure Answer Coverage, Not Citation Applause
The moment AI visibility became a topic, everyone started counting citations and celebrating screenshots of themselves in a ChatGPT answer. That is a vanity metric wearing a strategy costume. Hunt calls the order of operations out directly, and it is the discipline most brands skip.
Answer coverage is a number you can actually track. Take the master list you harvested from those four sources. Count how many of those questions have a published, findable answer on your site. That percentage is your real score. Coverage first, speed second, applause never.
The one leading indicator worth watching, once coverage climbs, is whether being cited actually earns anything. Seer Interactive's 2026 update, reported by SQ Magazine on May 7, 2026, found that cited brands earn roughly 120% more organic clicks per impression than uncited brands on AI Overview queries. So citations are not worthless. They are worth chasing after you have published enough decision knowledge to deserve them, which is the whole sequence Hunt is arguing for.
If you want this handled without building the spreadsheet yourself, the delegation ask is simple. Tell your agency: "Show me our answer-coverage score against the questions from our sales calls, support tickets, site search, and quote forms, and show me the plan to close the gaps." If they answer with a keyword report instead, you have learned something about your agency.
Frequently Asked Questions
How do I know which questions are worth publishing on my site?
Stop guessing and start harvesting. The questions worth publishing are the ones buyers already ask you before they purchase, and they live in four places you already own: your sales-call objections, your support tickets and DMs, your on-site search box, and your quote-form or configurator. Read the last 90 days of each, and every question that shows up more than a couple of times is a page. You are not inventing a content calendar, you are transcribing the answers you already give out loud.
Isn't this just adding an FAQ page?
No, and the difference is what makes it work. An FAQ page usually answers logistics with one-word replies, like the one-word "Yes" that Bill Hunt described in Search Engine Journal, which shut down a question more than 100,000 people had searched. Decision content answers the reasoning behind a purchase: will it fit my case, is it covered, how fast, how does it compare. Each of those deserves a full, honest answer with the caveats included, not a single line buried in an accordion. It is a different job with a different depth.
Will answering these questions online cost me sales calls?
It does the opposite. People who read a clear, honest answer to their real objection arrive at the call already sold on the fit, so the call gets shorter and closes higher. The buyer who reads an honest "no, this isn't right for your use" was never going to buy anyway, and now they don't clog your pipeline or return the product. Clear answers pre-qualify; they don't cannibalize.
Go look at your own product pages right now, on your phone, the way a buyer would. Count how many of them tell someone why to pick you instead of what you sell. If the honest answer is "none of them," you have just found the reason an AI keeps recommending the store that resells you. The good news is that the questions are already sitting in your inbox and your search logs, unanswered in public. If you want a second set of eyes on which gaps are costing you the most, book an AI visibility call and we will read your logs together. The strange part is that the most valuable pages on your future website are conversations you have already had.
