Every Monday morning my team runs the same set of buying questions through ChatGPT, Perplexity and Gemini. Questions a real buyer would ask: which platform fits a mid-sized logistics company, who provides this service in Israel, what does a solution like this cost. We log which companies get named in the answers, which sources the engines lean on, and what changed since last week.Â
We have been doing this every week for over half a year, first for ourselves and then for clients. The log tells a story that most B2B marketing teams have not caught up with yet.
THE FIRST MEETING MOVED INSIDE THE ANSWER
The story is this: the first meeting between a buyer and a vendor no longer happens on a website, or even on a search results page. It happens inside an answer. LinkedIn’s guide for B2B marketers (https://business.linkedin.com/advertise/resources/unlocking-ai-visibility) puts the number at 94 percent of decision makers already using generative AI during vendor research, and cites Gartner’s projection that organic search traffic will fall by half by 2028.
Whatever your feelings about those projections, the direction is not in dispute. The buyers in our client categories confirm it in almost every sales call: “we asked ChatGPT before we contacted you” has become an ordinary sentence.
What surprised me was not the shift itself. It was how mechanical the answers turn out to be once you track them long enough.
PATTERN ONE: ENGINES REPEAT WHAT THEY CAN VERIFY
When we placed one consistent, factual description of a company on its own site, in a press release and in a business directory, the same sentences started surfacing, sometimes verbatim, in AI-generated summaries within days. Not because anyone gamed anything. The model simply found three independent sources agreeing on who the company is and decided the claim was safe to repeat.
Companies whose description changes from page to page do not get that treatment. They get vagueness, or worse, they get skipped.
PATTERN TWO: “BEST OF” QUESTIONS ARE ANSWERED FROM DIRECTORIES
Ask any engine for the best agency, the best HR platform, the best anything in a given country, and the citations trace back to a handful of review directories and editorial roundups. A company can publish excellent content for a year and stay invisible on exactly the questions that carry purchase intent, because those questions are answered from sources it never bothered to appear in.
We watched competitors with thinner websites outrank clients in AI answers purely because a directory profile with four reviews existed and theirs did not.
PATTERN THREE: THE FEEDBACK LOOP IS FAST
LinkedIn’s research measured a median of under seven days from publishing a page to seeing it cited in AI answers, and our log agrees. This is the opposite of classic SEO, where you wait a quarter to learn whether anything worked. A factual page that answers a real buying question can show up in answers the same week.
So can a mistake.
THEN THE ADS ARRIVED
In September, OpenAI opened ChatGPT’s ad platform to advertisers in Israel, my home market after rolling it out across the US, Europe and Asia earlier in the year. The mechanics deserve attention because they break the habits of everyone raised on keywords. There are no keywords: you write a short free-text description of the conversations your ad belongs in, up to 280 characters, and the model decides the rest. There are no images yet.
Entire categories are also locked out: finance, health, legal, gambling, dating, politics. In a startup ecosystem heavy on fintech and healthtech, that detail matters more than the launch itself, and it rarely makes the coverage.
The ads only reach free-tier users. That is still an enormous audience, roughly 95 percent of a user base OpenAI reports at close to a billion people. But every paying subscriber, which in B2B tends to mean exactly the senior people you want, will never see a sponsored card. The only way to reach them inside the answer is to be part of the answer.
THE PAID LAYER IS THE SMALLER HALF
Ads switch off when the budget does. Presence inside the answer itself, earned through consistent facts, citable pages and third-party sources, keeps working across every engine, every pricing tier, and every category the ad platform refuses to serve.
WHAT I WOULD DO THIS QUARTER
If I ran marketing at a B2B company right now, I would start by asking the engines about my own category, this week, and writing down who gets named. Most teams have never done this and the result is usually uncomfortable.
I would then fix the boring things first: one factual company description repeated everywhere, real answers to the questions buyers actually ask, a presence in the two or three directories the engines keep citing. Only then would I think about the ad platform, and only if my category is allowed in.
None of this is exotic. It is closer to public relations than to growth hacking: decide what is true about your company, say it consistently, and make sure the sources the machines trust say it too. The machines are already answering questions about you either way. The only choice is whether you show up prepared.
BIO:
Erez Menashe is the founder and CEO of Crown, a global marketing partner for B2B technology startups, based in Tel Aviv. His team publishes weekly measurements of how AI engines answer B2B buying questions.
