“The Brands Falling Behind Aren't the Ones Underinvesting in AI Ads”, Similarweb's Liat Kidron-Sagiv on the Real Cost of Waiting

When ChatGPT ads launched, Liat Kidron-Sagiv, Senior Product Marketing Manager at Similarweb, says her team had data from day one, and what made it useful wasn't speed alone. It was the source: real user behaviour rather than constructed queries.

That distinction matters because AI ad placements don't trigger the way search or social ads do; they depend on intent building across a real conversation, something synthetic prompts can't replicate.

Liat has built her career around what she calls solving a “black box”: finding the moment a persona is flying blind, whether that's a demand gen manager, an SEO lead, or an advertiser trying to prove AI ads are real, and closing that gap with data instead of guesswork. In this conversation, she explains why she deliberately avoided promising clean, closed-loop attribution for AI ads before the rigor was there to back it up, what changed once she shared a stage with an actual top-3 ChatGPT advertiser, and why she believes the real risk for brands isn't a missed channel, it's still measuring organic, paid, and AI as separate decks instead of one connected journey.

You moved from partnership marketing to product marketing at Bringg and then to Similarweb, was that a deliberate pivot toward product-led positioning, or did the opportunity find you?

Bringg was a much smaller company, a startup. Their product supports last mile logistics, everything from fleet management and delivery orchestration to consumer communications. But the truth is a company like Bringg depends heavily on its partners. In e-commerce, delivery is just the last step, but all the systems before it need to work together for the whole flow to function, from discovery to the moment you hit purchase to the moment the package shows up. So partner marketing there was really a revenue led role. A lot of what I did was scoping joint value propositions and partner go to market strategies. After a while it was obvious I was already doing product marketing work alongside it, so the move felt natural, not like a leap.

Coming to Similarweb was more intentional. In this AI era I wanted to be at a data led company, because I believe that data is the new fuel of the tech world (and beyond). I don't actually think of Similarweb as product led, I think of it as data led. The real question is how our data helps customers make better decisions, in marketing, content, strategy, sales, retail, you name it. How they get that data varies too. Some want it visualized in our platform, but more and more want it through MCP, API, or a custom data feed.

I've always gravitated toward value led positioning, so moving to a company where the data itself is the value made complete sense to me.

If you had to explain what "solving a black box problem" actually means as a product marketing discipline, not as a company mission statement, but as a day-to-day muscle, what would you tell someone earlier in their career trying to build that skill?

Solving a black box starts with understanding what the product actually solves, and why that matters.

As a Senior PMM, I work closely with product, sales, customer success, marketing, analysts, and the market, so I understand the real pain points of the persona I'm marketing to. The way I think about it, a black box is when a persona is missing the information they need to do their job well. That's the muscle I'd tell someone earlier in their career to build: sit close enough to a persona that you feel the gap yourself, rather than assuming it. At Similarweb, my world is Growth, so I stay close to demand gen, PPC, copywriters, and the SEO and GEO teams. That's where the gaps actually show up (for me right now, this changes depending on the product you’re marketing of course!

A black box is never just a day to day inconvenience. It flows upstream. When a persona lacks the right data, it becomes harder to justify budget, defend initiatives, or prove value. The bigger the gap, the harder that case becomes to make. Giving people the right data closes that gap and turns them into champions who can make the case themselves. At Similarweb, that means arming our end users with the data to be excellent at their job, replace guesswork with answers, and prove ROI with confidence.

The discipline itself travels anywhere: find where a persona is flying blind, understand what that costs them upstream, and translate that gap into something both the product and the market can act on.

You called the AI Ads launch something you'd "personally been very excited about", what made this one feel different from other product launches you've worked on?

This was the first launch I've worked on that was tied so tightly to a brand new trend as it was happening. ChatGPT ads launched and Similarweb had data from day one, and what made that data actually differentiated is that it comes from real user behaviour, not synthetic prompts. That distinction matters more than people realize. A lot of AI visibility tools submit constructed queries and record what comes back, but that's not how ChatGPT ads actually get triggered, it depends on intent building up across a real conversation. Watching that unfold with real behavioural data, in real time, and being first to make sense of it for advertisers, that's what made this launch feel different from anything I'd worked on before.

You've posted about AI zero-clicks dramatically impacting organic traffic, for a brand that's underinvested in AI ad visibility, what's the cost of waiting another six months? 

This really changes between industries, and depends entirely on the organization. But what our own data keeps showing us is that AI visibility can't be optimized in isolation. The impact is downstream, a bit like brand awareness initiatives that are hard to measure one to one, but where the attribution is clearly correlated for brands investing across channels, organic and paid alike.

The value of an AI mention or an AI ad isn't always realized inside that single interaction. Sometimes we see referral traffic or a strong CTR right away, but a lot of the time it shows up later, in a branded search, a direct visit, a higher converting session. If you're only measuring the click inside the AI surface itself, you're missing a big chunk of where the return actually lands.

So instead of asking what a brand loses by waiting six months on AI ad visibility alone, I'd reframe the question: are they even set up to see the full journey, across organic, paid, and AI, as one connected path rather than separate team decks. The brands falling behind aren't just the ones under investing in AI ads, they're the ones still measuring each channel in isolation and missing the compounding effect of the middle of the funnel. That's the real cost of waiting: not a missed channel, but a blind spot in how ROI gets proven at all.

Doing the Natural Intelligence webinar meant sharing the stage with an actual top-3 ChatGPT advertiser, what did that conversation change about how you talk to prospects who are still sceptical AI ads are "real" yet?

What stuck with me most was how fast the conversation turned practical. When you're next to someone actually spending real money on ChatGPT ads, nobody's debating whether the channel is real anymore. The question becomes how do you do this well, and how Similarweb data supports that success. My role isn't to tell organizations whether they should advertise there or not. It's to help show who's there, what's working, what competitors are doing, and how it connects to or completes their other channels.

The partnership with Natural Intelligence wasn't just the webinar either. They were genuinely a design partner in helping us understand what metrics actually matter to advertisers, and that's what turned the whole conversation, and the product, from theory into practice.

Now I spend a lot of my time with customers, prospects, and sales reps walking through the next layer of questions: which prompts are owned at the topic or intent level, who owns share of voice, how landing pages are built, what creatives are winning, who you're up against, how you know it's working. The black box feeling around AI ads comes from these gaps, from not being able to see into the channel yet. Once people see the data, their confidence rises and they're able to make better, data driven calls.

As the person owning positioning for Ad Intelligence, what's a claim about AI Ads visibility you've deliberately avoided making until the data backed it up?

The one I was most careful about was suggesting we could offer clean, closed loop attribution for AI ads the way marketers are used to from search or social. Measurement here is still genuinely early. The models and surfaces are shifting month to month, so I didn't want to promise we could tell someone exactly what an AI ad impression turned into before we had the rigor to prove it.

What I was comfortable saying early on was narrower: we can show creatives, landing pages, topics, intent, share of voice, and more inside these new AI surfaces as they emerge. That claim holds up. The fuller attribution story is something I want us to earn, not get ahead of. In the meantime, organizations can start thinking about how they'd build their own attribution models using our data.

One thing that's made this feel real rather than theoretical is what we're already seeing in the data. Over 65 percent of ChatGPT advertisers are running AI specific campaigns with dedicated landing pages and their own tracking. That tells me the market isn't waiting around for perfect attribution either, they're building the infrastructure now, and we can help them do that with real visibility instead of guesswork.

If you were advising us on where to start be it search/display/social intelligence, brand protection, or the new AI Ads capability, given where the market's headed, what would you tell us not to skip?

I'd steer away from ranking one channel above the others, because they're not really competing. They're different angles on the same question, which is where your audience's attention is going and who's fighting for it there.

If I had to give real advice, it's that paid marketing doesn't live in a vacuum. An organization needs to look at how it actually acquires customers. Different models have different best practices and tactics, is this a B2B, B2C, D2C business or something else? That changes the answer. Are the campaigns built for awareness or for conversion? That changes it too. You need to look at the whole marketing strategy, organic, referral, email, and everything in between, and then make the calls that are right for you.

For some companies that still means core paid search and social. For others it might be something else entirely. And if you're already seeing even a small amount of traffic from AI surfaces, that's worth doubling down on.

But the truth is, this is very individual, and I'd be cautious before handing out one size fits all advice.