Back in April, we published our early take on HubSpot’s AEO tool — what it did well, what it didn’t, and what we were still figuring out. That post got picked up in HubSpot’s weekly tips roundup, which told us we weren’t the only ones trying to make sense of this.
Two months later, we have actual numbers. More importantly, we have a closed deal.
A prospect reached out a few weeks ago and told us they’d been “recommended by a very reliable source.” We assumed it was a referral partner. It was AI. They’d asked their LLM who to talk to about HubSpot implementation, our name came back, and they took it seriously enough to book a call. That deal has since closed and is one of the largest projects we’ve landed in several months.
It’s not the only one. In a single week we quoted roughly $120,000 worth of work sourced directly from AI referral traffic. Some of those prospects cited specific blog posts we’d built with the AEO tool’s help by name, before we even brought the posts up. That’s the part that matters most to us: this isn’t just a visibility metric going up on a dashboard. It’s showing up as SQLs who already trust us before the first call.
Here’s what we tracked, what moved, and what we changed to make it move.
Brand Visibility: 4% to 17%
We set up HubSpot’s AEO tool on April 22nd. At that point, our brand visibility score sat at 4%. By June 30th, it was 17%.
Obviously not a mission accomplished moment, but we went from almost invisible to gaining real traction.
We’d already done a fair amount of on-site AEO groundwork before that — structured data, schema markup, the usual technical foundation. The 4% starting point reflects where we stood after that work, which makes the climb to 17% even more notable. The tool isn’t just measuring whether you’ve checked technical boxes. It’s measuring whether you’re actually getting picked.
That climb hasn’t been a straight line, either. At several points we’ve overtaken some of the top HubSpot Solutions Partners in our category for specific prompts — and then lost that ground a few weeks later when someone else published something stronger. Visibility in AI search isn’t a “set it and forget it” SEO ranking. It moves week to week, sometimes day to day. If you’re going into this expecting a permanent leaderboard position, you’re going to be disappointed. If you’re going into it expecting an ongoing fight for attention, you’ll be a lot more prepared for what it actually takes.
Steady Brand Sentiment
Fortunately, Simple Machines’ brand sentiment score has been healthy at around 70% since we started tracking. There has not been much change here, but this makes sense as our efforts so far have been focused on improving visibility. More to come on this down the road when we experiment more with tactics to improve our score.
The Platforms Aren’t Treating Us the Same
This is the part of the data that surprised us most: our visibility isn’t even close to consistent across AI platforms.
We’re sitting at over 40% visibility on Gemini. ChatGPT and Perplexity are fluctuating somewhere between 5% and 25%, depending on the week. (Still waiting on Claude, but we’ve been told by HubSpot it’s coming.)
We don’t have a confirmed reason for this — none of the platforms tell you how they weight or source content, so anyone claiming certainty here is guessing. Our working theory is that Gemini is responding more strongly to structured, well-tagged content than the other two. We can’t prove that’s the mechanism. What we can say is that the gap is real, it’s been consistent over two months, and it’s big enough that if you’re only checking your average visibility score, you’re missing the more useful story underneath it.
If you’re investing in this, don’t just track your overall number. Break it out by platform. The platform that’s ignoring you today might be the one your next ICP segment uses.
What We Changed
The AEO tool will hand you prompt recommendations and even help you generate content against them. Very useful, but not enough on its own.
If you take the tool’s recommendation and ship it close to as-is, you might see a short-term bump — but so will everyone else doing the same thing with the same tool. That’s basically a coin flip on whose content the algorithm happens to like better this week.
What’s worked for us is using the tool to identify which prompts matter, then doing real work on top of that before anything gets published. Concretely: we run the target prompt through Claude and ask for ten different angles on it, then pick the three that are most differentiated and most ours — built around our actual point of view, our own client work, proof we can back up. The prompt recommendation tells us where to aim. It doesn’t write the thing that gets cited.
That’s the piece we’d tell anyone starting this to take seriously. The tool is infrastructure, not strategy. The content still has to sound like you, prove something you’ve actually done, and say something your competitors aren’t already saying in their own AEO-optimized posts.
AI Referral Traffic: Smaller Than You’d Expect, But the Trend Line Matters More
Here’s the number that’s easiest to misread: AI referral traffic to our site is up 12% since we launched the tool on April 22nd. That’s a real increase, but on its own it’s not a headline.
The more telling number is this: AI referral traffic has doubled since January, while organic search traffic over that same window has dropped. We’re not claiming a traffic explosion. We’re pointing at a shift in how people are finding us — and it lines up with what we’re seeing anecdotally on the sales side, where prospects are showing up already informed, already trusting us, sometimes citing specific posts before we’ve mentioned them.
A modest increase in raw traffic next to a meaningful change in where that traffic originates is, to us, a more honest and more useful data point than a single inflated growth percentage would be.
Where Things Go From Here
Visibility moves fast. Platforms behave differently. The tool’s own recommendations get less effective the more people copy them verbatim. None of that means AEO isn’t working — our numbers say otherwise — but it does mean it’s not a campaign you launch and walk away from. It’s closer to an ongoing practice, the same way SEO was a decade ago, before everyone treated it like a checklist.
We’re going to dig into the specific gotchas — the things the tool’s reporting doesn’t catch, where it lags, where it’s flat-out wrong — in a follow-up post. For now, if you’re earlier in this process than we were, go back and read where we started. The early numbers looked nothing like these. In a couple months from now, it will likely be a different story again.
