What does generative engine optimization actually change for a B2B company? Elad Hefetz has been building websites for enterprise tech since 2014, first as CodersClan, now as Airfleet, a 55-person team spread across a dozen countries. In Episode 2 of the NoRobots Podcast, Vlad sits down with Elad to talk about GEO, why AI recommendations rarely come with a link, and the two pieces of advice he thinks marketers should ignore.
Finding the niche by accident
Airfleet started as CodersClan, a paid Q&A platform for developers helping developers. But the people signing up weren’t developers. They were marketers and founders who needed technical help their internal R&D teams were too busy to give.
Marketing teams wanted a landing page, a chatbot, a connection to their automation stack. Engineering was heads-down on the product. That gap turned into a business, and by now the focus is narrow: enterprise tech B2B, where a purchase involves a buying committee of six to ten people and the website has to serve every one of them.
Saying no to money
Narrowing down meant turning away work. Elad is direct about how that felt: the hard part was “saying no to people that are willing to pay you money,” especially when you technically could help them.
What made it easier was thinking in opportunity cost. Take an okay client and you may not be free when a great one shows up. It’s the same instinct Josh Webber described in Episode 1 when he talked about earning the right to say no.
Why WordPress won for them
Airfleet tried plenty of platforms. Around 2020 they bet on Gatsby.js, built their own platform on top of it, and six months later admitted it wasn’t going to work. They migrated seven clients back to WordPress, some of them for free.
The reasoning now is less about technology and more about risk. WordPress is mature, extensible, and there’s no vendor holding the pricing lever. If a hosted builder doubles its price tomorrow, its customers have no move. Next.js with a headless CMS is a fine choice, Elad says, but switching would mean rebuilding years of accumulated capability for no client benefit.
When the package doesn’t fit
Fixed packages work for defined projects: a website, a visual identity, an SEO or GEO engagement. But scope creeps, and some work has no knowable scope at the start.
Their answer is to change the shape of the deal rather than argue about it. Ad hoc projects, hourly work, or an hourly subscription with a flat monthly fee that gives the client access to everything the agency does. Different problems, different contracts.
GEO starts with prompts, not rankings
This is the part most teams get wrong. Everyone wants to appear in ChatGPT. Elad’s reframe: “It’s about appearing for the right audience for the right intent.”
Which means the first task isn’t optimization at all. It’s working out what your buyers actually type into these tools. Airfleet built its own AI visibility platform to track unlimited prompts per client, because off-the-shelf tools cap you at a package. Once you know the prompts, you can see which sources the models lean on, and those sources vary by industry. In cybersecurity it might be a handful of publications, specific Reddit threads, or YouTube.
Mentions are the new backlinks
The structure of GEO mirrors SEO more than people expect. Technical setup, content, third parties. The difference is in the third bucket: “instead of backlinks, it’s about mentions.”
And GEO is influenced more by what’s written about you elsewhere than by your own site, though your site still shapes the sentiment of the answer. The two disciplines aren’t alternatives. As Elad puts it, “You can’t do just SEO or just GEO. It doesn’t exist.”
The shortcut, if there is one: with backlinks, any solid authority site helps. With GEO, the same specific sites keep showing up as sources. Get mentioned there and a citation is close to guaranteed.
AI recommends you. Nobody clicks.
Here’s the measurement trap. Ask an AI for the best agencies in a category and it names three. It usually doesn’t link any of them. So the buyer copies the name, pastes it into Google, and arrives as a branded search.
Which means a working GEO strategy shows up as rising brand searches, more direct traffic, and more LinkedIn activity, not as a traffic spike from chatgpt.com. Judge it by referral traffic alone and you’ll conclude it isn’t working.
On lead quality, Elad hasn’t seen a durable difference. Early on, AI-sourced buyers seemed more educated, but that’s evening out. What has clearly increased is bot traffic and automated spam submissions, which are hard to filter precisely because they look like real browsers.
Two pieces of advice to ignore
First, the AI content pendulum. It swung from “AI content can’t rank” to “automate everything,” and Elad thinks both are wrong: “The truth is somewhere in the middle.” Automate the production work, but publish an actual point of view. Fully automated content sites do show climbing impressions in Search Console, then fall off a cliff.
Second, Wikipedia. Getting listed there is roughly ten times harder than landing on a listicle, and it mostly feeds top-of-funnel prompts where the AI answers the question without naming your brand. His verdict: “it’s not a good use of your time and money.” An ordinary industry listicle does more.
Tel Aviv hours, US clients
Most of Airfleet’s clients are in the US, so Elad runs “two shifts” — one before the kids go to bed and one after. The team is global: 55 people, only nine in Israel, hired wherever the talent is, and almost always full-time rather than per project.
The cultural gap took adjusting to. “We say what we think, we’re very direct” doesn’t always land with US buyers, and he had to learn to frame the same point differently. The upside is that American clients tend to value good service and pay for it, which raises the standard for everything the agency does.
Find Elad at airfleet.co or on LinkedIn, and watch the full episode on YouTube.
