Nine thousand two hundred and eighty-eight citations. Twenty-one of them ours.
That's the last 30 days of our own AI search visibility, measured in our own HubSpot portal, across the 25 prompts we track. We sell AI-readiness work. Those are our numbers. I'd rather publish them than describe someone else's.
Twenty-five prompts, scoped to the United States, weighted to the things we actually sell: CRM migration, integration, HubSpot consulting, revenue operations, and AI and Clay work. We rebuilt the set this month, because the version before it tracked generic "RevOps" category questions that nobody types and two branded prompts that only ever returned us.
HubSpot ran those 25 prompts against ChatGPT, Gemini and Perplexity. 972 answers came back. Here's what they contained.
We were mentioned in 100 of 972 answers. Roughly one in ten. The average answer named 1.9 competitors.
Those answers cited 9,288 sources. Twenty-one were on our domain. That is 0.2% of the citations in answers to questions about work we do every day.
Across those same answers, brand mentions broke down like this. RevPartners 9.2%. Aptitude 8 5.1%. New Breed 5.0%. SmartBug Media 4.9%. Huble 4.5%. Simple Machines Marketing 4.1%. Six & Flow 1.9%. Evenbound 1.5%. Carabiner Group 0.9%.
Penguin Strategies: 0.06%.
One caveat that matters, and it cuts both ways. This is share of voice inside our tracked prompt set, not the whole internet. We chose these 25 questions because they are the ones we want to win. So the number is not "how visible is Penguin generally." It is "on the exact questions we care most about, here is where we sit." That is the more useful and the more uncomfortable version.
The engines disagree about us. ChatGPT mentioned us in 51 of 324 runs, about one in six. Perplexity, 30 of 324. Gemini, 19 of 324, roughly one in seventeen. Same company, same prompts, same month, three very different answers. If you are measuring AI visibility as a single number you are averaging away the finding.
Perplexity cites the most and cites us least. It produced 6,059 of the 9,288 citations, nearly two thirds. Four of them were ours. An engine that is hungry for sources and still not reaching for yours is telling you something specific about your content, not about your brand.
hubspot.com is 31% of every citation in the set. 2,880 of 9,288. When someone asks an AI about HubSpot migration or HubSpot integration, the answer is mostly built out of HubSpot's own documentation, and then a long tail of agency blogs. Which means the realistic goal is not to outrank HubSpot. It is to be the agency in the tail that gets named.
And the tail is winnable. A B2B agency blog we had never benchmarked against showed up 542 times. Others in the 100-to-200 range. These are not enormous brands. They are sites that answered one narrow question well, repeatedly.
The temptation after seeing numbers like these is to publish more. We did the boring thing first.
We wrote one canonical description of what Penguin is, and put the identical wording on every public surface: "B2B revenue operations consultancy and HubSpot Elite Solution Partner. CRM implementation, migration, integration and ongoing optimization for B2B technology, manufacturing and professional services companies."
Then we went and fixed the places that disagreed with it. Our founding year was wrong in two public places. Our partner tier was listed as a tier we left behind. Several profiles still described us as a marketing agency, which we have not been for years. Spelling was inconsistent across surfaces, which sounds trivial and is exactly the kind of thing that makes an entity look like two entities.
None of that is content marketing. It is making sure that when a model assembles an answer about us, the sources agree with each other. An AI model that finds three different descriptions of your company does not average them. It picks one, or it picks someone else.
Then we narrowed the topics. "HubSpot implementation" is a fight with every partner on earth and we will not win it this year. "Salesforce to HubSpot migration for manufacturers" is a question we can answer better than anyone, and there are not 400 agencies competing for it.
AEO is the quietest item in HubSpot's Fall release and the one I'd point marketing leaders at first. Every other Marketing Hub feature in the release produces content faster. AEO is the only one that tells you which content to produce, and then measures whether it worked.
It's included with Marketing Hub Professional and above, no credits.
Two things changed at Unbound.
AEO now drives Campaign Agent directly. When the dashboard flags weak topical authority, one click takes you into Marketing Studio, where Campaign Agent reads the recommendation, pulls your brand identity and ICP from Context Home, and proposes a connected campaign plan. Content Agent builds the assets. AEO re-measures citation rates and feeds the next recommendation. HubSpot calls this the insight-based activation pattern.
And there is now a public AI Search Sensor at hubspot.com/ai-search-sensor, which partners can embed. It's a quick read on how visible a company is in AI answers. We've put it on this page so you can run yours without leaving.
AEO surfaces the gap. Campaign Agent proposes the plan. You review and edit the channel mix. Content Agent writes and self-grades. Nurture Agent personalises the conversion emails. AEO re-measures.
Five of six steps are automatable. The channel-mix review is not, and HubSpot says so itself. If a campaign runs from AEO insight to published content without a person deciding where it should appear and why, you have automated the production of content nobody asked for. Given what the citation data says about which sources get picked, that is a real risk.
Citation rate on the three to five topics you actually want to own, today, before a single campaign runs. Which engines cite you and which don't, separately, not averaged. Which domains are cited where you aren't, including the small ones. Then after a cycle: the same numbers, plus whether any pipeline traces back.
Scope prompts to the geography you sell in. Delete the branded prompts; a question with your name in it will always return you and teaches you nothing.
And be patient with it. Citation is a lagging indicator. Expect a quarter, not a fortnight.
AEO is a Marketing Hub Pro+ feature, so Starter and Sales-only portals don't get the dashboard. The loop assumes your Context Home is right; an AEO campaign built on a stale ICP produces authoritative content aimed at the wrong buyer. And prompt capacity is 25, which forces a useful discipline: you have to decide which 25 questions matter.
We'll publish these numbers again in 90 days, same 25 prompts, and show whether any of it moved.
Run the sensor on your own company above. If the result stings and you want to talk about what to do first, our pricing page shows how we work.
Part of our guide to the HubSpot Unbound 2026 release: everything HubSpot announced, on one page.