Try this. Open ChatGPT or Claude and ask it to compare your product against your biggest competitor — pricing, weaknesses, real customer complaints, the works. Read what comes back.

This is exactly what your customers are doing before they land on your site. Thanks to GenAI, every buyer has a personal analyst helping them make smarter buying decisions.

Hashi's take: this is not a threat to manage. It's a flywheel — and the businesses that feed it are going to pull away from the ones that starve it.

If your marketing lead still treats AI search as a curiosity, forward this their way.

Gartner surveyed 645 B2B buyers and published the results in May. Forty-five percent used GenAI during a recent purchase. And 69% now use sales reps to validate what the AI already told them.

Validate. Not learn — validate. The customer arrives with a formed opinion, and your rep's new job is confirming or correcting it.

The questions themselves have changed too. When researchers analyzed 8,500+ ChatGPT prompts, the searches it ran for users were dominated by three words: "reviews," "comparison," and "features." Nobody types keywords at an AI. They ask full questions, then follow up, then ask the question behind the question. Adobe found the same pattern on the consumer side: people are 83% more likely to use AI for complex purchases. The bigger the decision, the more homework gets done.

So the customer walks in prepared. Just not always correctly.

AI makes things up. Fluently, in complete sentences, with no tell in its output that it's made it all up.

The EBU and BBC ran the largest study yet on this — over 3,000 responses from ChatGPT, Copilot, Gemini and Perplexity, across 18 countries and 14 languages. Forty-five percent of the answers had at least one significant problem. One in five had real accuracy errors: invented details, outdated information stated as current.

That was news, a subject with enormous amounts of published material to draw on. Your product catalogue is a much thinner slice of the internet. The odds of an AI filling a gap with something it made up go up, not down.

So a second kind of buyer shows up. Informed and wrong. Convinced your service includes something it doesn't. Quoting a price you retired two years ago. Comparing you to a competitor on a spec neither of you publishes. And because a machine delivered it in a calm, confident paragraph, they trust it more than they'd have trusted your salesperson.

Sharper questions from one group. Confident nonsense from the other.

How do you fight both? Three ways.

Until a few years ago you could run a business on thin content. A glossy page, three benefit bullets, "book a demo to learn more." It worked because the customer had no way to check. The information lived with you, and they had to come get it.

When a customer asks an AI whether your product handles their specific use case, it answers from whatever depth exists — your spec sheets and honest FAQs if you published them, a competitor's comparison page or a Reddit thread if you didn't. And when there's nothing solid anywhere, it fills the gap with something invented. Every question your content can't answer gets answered by someone else, or by nobody, badly.

Most businesses have this exactly backwards. Adobe found that homepages average 75% content visibility to AI tools, while product pages — where buying decisions actually happen — score just 66%. The page that matters most is the one AI can read least.

This is AEO and GEO, which I wrote about in May. Back then getting cited by AI engines looked like a marketing tactic. It's bigger than that now. Being cited isn't worth much if what gets cited is thin, and publishing the real answer is the only defence against the made-up one that actually scales.

Depth means more than most companies expect. Current pricing logic. What your product can't do. Honest comparisons against the competitor you'd rather not name. Marketing teams have spent their careers withholding exactly this material and legal will flinch at some of it, which is why this is a decision for whoever is reading this, not for the SEO agency.

That handles the buyer before they ever contact you. It does nothing for the one already convinced of something wrong.

Go back to that Gartner number. Sixty-nine percent of buyers now use reps to validate what AI already told them. That's the job description now. Not informing. Checking.

That's a different skill, and almost no sales script was written for it. Checking often means correcting, and correcting a confident person is nothing like explaining something to someone who walked in blank. Do it badly and the buyer digs in, defends whatever the machine told them, and leaves. This is where I'd spend real training hours. It's the cheapest of the three and probably the highest return.

Start by finding out what your reps are up against. Ask ChatGPT, Claude, Gemini and Perplexity about your pricing, your limitations, how you compare to the competition. I'd bet most leaders haven't looked yet. Then build the retraining around what comes back — how to handle the sharper objections, and what to do when the machine simply got it wrong.

Your reps can only fix the conversations that reach them.

Better-informed customers ask more questions, not fewer. More specific ones, more often, and most of them before anyone from your team is involved. You can't staff your way through that.

This is where AI creates real value in a business. Most AI initiatives fail that test — they automate something nobody was struggling with. This one doesn't. It absorbs work that's genuinely growing, it runs on your own data instead of someone else's model of the world, and it gets better every month because your customers keep feeding it.

A triage layer answers what your published content already covers, routes the genuinely hard ones to your people with full context attached, and keeps a log of every question it couldn't answer well and every wrong assumption it had to correct. Keep the bar high — 68% of consumers expect a chatbot to match a skilled human agent — because a bad bot makes things worse, not better.

But that log? That log is next month's content plan, written by your own customers.

And that's the moment those three stop being a list and start being a circle. You publish depth. AI engines cite you, because you're the deepest source on your own product. Better-informed customers arrive with harder questions. Triage catches the ones you haven't answered and hands them back to marketing. You publish those. The wheel turns again, and every rotation makes the next one cheaper.

Everything about how you sell assumes you know more than the buyer does. The ads, the scripts, the support playbook. That was true for a long time. It isn't anymore.

AI is mass-producing expert customers, and most businesses are still set up to sell to amateurs.

None of this is hard to fix. It's a shift in thinking: reposition what you publish, prepare your people for a buyer who arrives already briefed, and use AI to handle what AI created. The volume and the misinformation both come from AI. So does the only tool that can keep up.

An informed customer used to be the salesperson's nightmare. I think that instinct has it backwards.

Information asymmetry protected bad products. If your product only sold because the customer didn't know better, AI is coming for you, and honestly, it should. But if what you sell holds up under scrutiny, an expert customer is the best customer you'll ever have. They arrive pre-qualified, they ask questions you can actually win on, and when they buy, they stay.

- Hashi