Devstars
HubSpot launched its Answer Engine Optimisation tool on 14 April 2026. Fifty dollars a month, 28-day free trial, no HubSpot subscription required. On its own, that would be interesting. What makes it genuinely useful is the research they published alongside it, based on millions of prompts across ChatGPT, Gemini, and Perplexity, with data from their XFunnel acquisition and a year inside their own business.
This is the most practical public GEO data set we have seen so far. Below is what actually stands out, what it means for the work we are doing with clients, and what to do with it this quarter.
HubSpot’s own customers are watching organic traffic fall 27% year on year. That is the Jaw Effect playing out in real numbers across hundreds of thousands of sites. Impressions holding up, clicks falling off, because the AI engine is answering the question before the user gets anywhere near a website.
The useful thing about HubSpot publishing this now is that it validates what we have been telling clients one to one for 18 months. The conversation is no longer “do you believe me?” It is “here is what HubSpot just proved.”
Everyone has been calling the blog dead for two years. It is not. It has changed jobs.
HubSpot’s analysis found that 62% of AI citations come from blog posts and listicles. Product pages barely feature. Their own log data tells the same story from the other side. Twenty per cent of bot traffic to hubspot.com lands on the blog, but those blog pages generate the 62% of citations. In other words, one in five bot visits is punching three times above its weight.
The shift in mindset matters more than the numbers. Aja Frost, who leads AEO at HubSpot, put it well. The blog used to be a traffic channel. It is now an influence channel. You cannot measure its success by direct visits any more, because the visits that count are bots reading the page and using the content to shape answers. You measure it by citations earned and by the downstream deals that close faster because an AI engine recommended you first.
Beeri Amiel, the XFunnel co-founder now running the tool inside HubSpot, called it “business to bot to consumer.” It is a clunky phrase but it is the right mental model. Your blog is talking to the bot, the bot is talking to the buyer, and the buyer is turning up on your site already convinced.
This is the one that will rattle any client still living in a pure SEO mindset.
HubSpot’s data shows that the correlation between ranking on Google and being cited in LLMs is weak, and the relationship varies by engine. On Gemini and Google’s AI mode it is reasonable, because those systems sit on top of Google’s own indexing. On ChatGPT there is almost an inverse relationship. The higher you rank on Google, the less likely ChatGPT is to cite you.
This makes sense if you think about user behaviour. The average prompt on ChatGPT is 23 to 25 words long, not three. People are describing very specific situations and asking very specific questions. ChatGPT is looking for content that matches that specificity, not content that ranked for a short-tail keyword five years ago.
The practical implication for your clients is direct. A site that ranks well on Google is no guarantee of AI visibility. You need to treat GEO as a separate discipline with separate measurement, and you cannot rely on your SEO agency to tell you how you are doing in ChatGPT.
When the AI engines are deciding whether to trust you, they are not just looking at your backlinks. They are looking at what other humans say about you. And three platforms play an outsized role in that decision.
The reasons are structural. Google owns YouTube. Microsoft owns LinkedIn and has the OpenAI relationship. Reddit has signed content partnerships with both Google and OpenAI. These three platforms have formal data pipes into the major engines. So the LLMs are weighting them heavily, and HubSpot’s data confirms it.
For most clients this changes the calculus on their social strategy. A proper LinkedIn cadence, a handful of YouTube explainer videos with clean transcripts, and strategic Reddit participation are no longer optional extras. They are the authenticity signal the engines are weighting most heavily.
What counts as a citation on each platform matters. On LinkedIn, it is a link to an article or post. On YouTube, the engines are pulling directly from transcripts, sometimes citing specific timestamps in the video. Which means your YouTube content needs to be searchable inside the video, not just findable as a whole asset. Captions, chapter markers, and answer-first phrasing inside the first minute all matter now.
This one is genuinely good news for small businesses and niche clients.
Because prompts are long and specific, the AI engine is looking for content that matches that specificity. You do not need to beat the biggest dogs in your category on broad keywords. You need to own a very specific corner of the answer space.
For a hair salon that means a page about late-night appointments in a named London borough with named stylists and actual opening hours. For a storage company it means a page about secure self storage in a specific location with real access hours, real security features, and real pricing. For a niche engineering client it might mean the only publicly indexed guide on repairing a specific component in a specific piece of equipment.
If you are a small business, you almost certainly have information that only you and maybe one other company in the world actually have. Put that information on the page in plain English, answer-first, and you will out-cite generic industry pages every time. The AI does not care about your domain authority in that niche. It cares whether you answered the question properly.
This is where the commercial story gets interesting.
HubSpot’s data shows that most AI citations disappear after six months, and a significant percentage change within a single month. Unlike SEO, where you can rank a page and coast for two or three years, GEO requires constant optimisation, refreshing, and production.
If your heart rate goes up reading that, good. That is the right reaction. It means you cannot set it and forget it. It also means that if your competitor takes their eyes off the ball for a quarter, you can walk in and take the citations. First-mover advantage is real here, and it compounds.
Commercially, this is why GEO belongs on a retainer, not as a one-off audit. Clients who buy a point-in-time report will be back to square one in six months.
A few specific places we are updating our approach this quarter.
For every existing client, we are running a GEO Readiness snapshot. Five to ten high-intent prompts across ChatGPT, Gemini, and Perplexity. Are they cited? If yes, from what sources? If not, who is being cited instead? This becomes the baseline, and we review it monthly. HubSpot’s free AEO Grader is a reasonable starting point if you want to see the mechanic in action. The proper tracking sits in the $50/month tool.
For every new pitch, we are leading with the Jaw Effect. Show them their own Google Search Console data with impressions rising and clicks falling. Then show them who is getting cited in AI instead. The cost of inaction is no longer abstract. It is the deal they just lost because an AI engine told the buyer to use a competitor.
For content production, we are shifting the blog emphasis. Listicles, comparison pieces, and how-to guides earn more citations than product pages. So the content calendar leans that way, with question-led H2s, answer-first opening paragraphs, and FAQ schema on every long-form piece. We have been moving this direction anyway. HubSpot’s data just validates it.
For platform strategy, we are writing a LinkedIn cadence into every Fractional CMO engagement, pushing YouTube for any client with a decent explainer story to tell, and flagging Reddit as an authenticity play for anyone who can credibly participate. The days of “LinkedIn nice to have” are over.
The $50/month tool is useful. The 28-day free trial is genuinely useful even if you have no intention of buying, because it gives you a baseline visibility score across ChatGPT, Gemini, and Perplexity without you having to manually prompt each engine.
The limitations are worth naming. It tracks those three engines, not Claude, not Grok, not AI mode separately. The recommendations are CRM-aware if you are on Marketing Hub Pro, and generic if you are on the standalone tool. The Marketing Hub integration is where the real value lives, and that is the $900/month HubSpot for Marketers bundle, not $50.
We are not planning to resell it. We are planning to use it as a research and baseline tool for clients, alongside our existing stack. If a client is already on HubSpot Marketing Hub, it is a no-brainer to switch it on. If they are not, we can get to a similar answer with manual probing and a spreadsheet, at the cost of a bit more of our time.
The biggest shift HubSpot has confirmed is the one we have been telling clients for a year. The blog is not dead, it has changed jobs, and the job is now to influence bots, not to drive traffic. Measure accordingly, budget accordingly, and the 12 to 18 month first-mover window is yours to take.
If you want to see where your brand currently shows up across the major AI engines, we run a GEO Readiness snapshot for a handful of new clients each month. Five to ten prompts, three engines, delivered as a short report with three prioritised actions.
Any questions, drop me a line.
Stuart Watkins is the founder of Devstars Jersey, a digital growth consultancy specialising in GEO, bespoke web software and fractional marketing direction. Devstars has been building websites and digital strategies for ambitious businesses since 2003, from offices in Jersey and London.
Tell me what you’re trying to fix. Half an hour, no pitch, no slide deck.
If we’re the right fit we’ll talk about what’s next. If we’re not, I’ll point you to someone who is.
