Across the 300,000+ HubSpot customers analyzed worldwide, organic traffic fell by 27% on average year over year. The figure comes from HubSpot itself, shared when it published its first-quarter 2026 results and repeated in the webinar we co-hosted with the vendor last June. This traffic has not disappeared. It has moved to ChatGPT, Gemini, Perplexity and other LLMs, where your buyers now ask their questions and build their shortlist without necessarily going through a Google results page.
Since spring 2026, HubSpot has included a dedicated tool in Marketing Hub to manage this visibility. We tested it on our own visibility, with a measurable result in three weeks.
This article explains what AEO is, why the topic has become urgent, how HubSpot's AEO tool works, and what our own experiment taught us. The figures quoted come from the Markentive × HubSpot webinar of June 17, 2026, unless stated otherwise.
AEO refers to all the practices that aim to improve a brand's visibility in the answers of AI engines: ChatGPT, Gemini, Perplexity, Claude, etc. You will also come across the acronyms GEO (Generative Engine Optimization) or Answer Engine Optimization. Three names, one reality: the counterpart of SEO for LLMs.
The fundamental difference with SEO lies in when everything is decided. On Google, the goal was to rank at the top of a results page that users would browse, compare and click through. On an LLM, there is no results page: there is an answer. If your brand appears in this first answer, you enter the conversation, then the buyer's benchmark. If it does not, you lose the prospect without even knowing it. There is no page 2.
Four figures shared by HubSpot during the webinar are enough to set the scene.
900 million weekly users on ChatGPT (early 2026 data), and around 750 million on Gemini. If you are still wondering whether your buyers use LLMs, this figure answers for you. The question in 2010 was “do we need a Facebook page”. The question in 2026 is “are we cited by ChatGPT”.
25% of searches worldwide now happen on LLMs. The calculation is worth breaking down: there is roughly one LLM query for every two queries on a traditional search engine, and nearly 60% of LLM queries are searches (the rest is writing, translation, code). That makes a quarter of all searches. Reached in three and a half years, whereas Google took nearly thirty years to build its dominance. Ignoring AEO today means giving up on optimizing your visibility on one search in four, on a trend that will not reverse.
Traffic from LLMs converts 4.4 times better than organic traffic, again according to HubSpot data. The logic is mechanical: the buyer has already conversed, compared and shortlisted in the LLM before clicking. They arrive on your site already self-qualified. HubSpot in fact treats this traffic as its main indicator of buying intent. Fewer visits, but visits that count: the funnel has narrowed at the top and widened at the bottom.
And the -27% organic traffic mentioned at the start completes the reasoning: early-journey traffic, the traffic SEO captured through volume, is the traffic LLMs absorb first.
Our view at Markentive: the current window looks like SEO in 2004. Few companies seriously optimize their AI visibility, the first-mover advantage is real, and it is closing. Waiting for AEO to be “mature” before investing is the surest way to relive SEO in 2015: more competitive, more expensive, slower.
Since 2026, Marketing Hub natively includes an AEO module: 25 tracked prompts in the Pro edition, 50 in Enterprise, with additional packs available. The “basic” setup takes about five minutes: domain, competitors, ICP, products offered. Then each prompt you declare is sent daily to ChatGPT, Perplexity and Gemini (Claude is announced for the end of 2026), and the answers are analyzed.
A configuration subtlety that often surprises people in workshops: a prompt corresponds to a market, and therefore to a country. A company present in France, the United Kingdom and Germany will have to configure three separate prompts to track the same question in its three markets. With a quota of 25 prompts in Pro, this needs to be budgeted from the start.
The tool answers three questions, in this order.
For each tracked prompt, the tool checks whether your brand is cited in the answer, day after day, then aggregates everything into an overall visibility score. During the webinar demonstration, the portal configured for HubSpot showed 81.05%: across all tracked prompts, the brand appeared in just over four answers out of five.
“The overall visibility score is a vanity metric if you stop there. Being pleased to see 80% or disappointed to see 15% does not trigger any action. The value starts when you break it down: by buyer journey stage, by persona, by prompt. That is when you discover that an entire stage of the journey, consideration for example, is underused, and you know what to produce.” Stéphane Duwicquet, Manager Business Unit Growth, Markentive
An absolute score does not say much. Is 40% visibility good or weak? Impossible to say on its own. The “competitor landscape” table translates this score into share of voice: how often you are cited compared with the competitors you have declared. A modest visibility score with a dominant share of voice is still a good position. The reverse is a warning sign. A visibility score looked at without its share of voice is a vanity metric, and we also tell this to our clients who show 80%.
This is the most actionable part, across two tabs. The citations tab breaks down the sources LLMs use to build their answers: which formats, which types of sites, which specific domains. Two lessons shared during the webinar are worth remembering. First, product pages, blog articles, rankings and comparisons account for about 70% of the formats cited. Second, the dominant sources are “peers” in your industry: sites in your ecosystem that are not your competitors (such as G2 in software), which opens up very concrete opportunities for editorial partnerships.
The recommendations tab turns these analyses into actions: articles to create (with a suggested title and a justification based on the visibility gaps found), Reddit threads to take part in (less common in France, but the logic applies to any industry community), source sites to contact. For portals with Content Hub, generating the recommended articles takes one click, and the result should then be treated as a first draft to enrich, never as publishable content.
On one central point, yes: the balance between authority and relevance is reversed.
Google ranked pages first by authority (backlinks, domain reputation), then by relevance. That made sense given the raw material: an average Google query is 5 words long, and with 5 words the engine has to guess the context. An average ChatGPT prompt is about 350 words long. The LLM has the persona, the industry, the use case, the budget constraint. It no longer needs to guess: it looks for the content that answers exactly. Relevance comes first.
The editorial consequence is direct: produce very specific content by target, product, industry and buying journey stage, rather than generic high-authority pages. Highly precise content on a narrow topic can beat a powerful but vague domain. This was rarely true in SEO. It is in AEO.
The other break is speed. HubSpot demonstrated this with its own pricing: LLMs were hallucinating about its prices and giving prospects wrong answers. Using the AEO tool, the team found that its blog was very frequently used as a source, published a highly structured article dedicated entirely to pricing, and saw the answers corrected within the following hours. About two hours. Where an SEO strategy is measured in months, a well-targeted AEO action can be measured in hours or days, because prompts are replayed daily.
This case illustrates a habit to adopt before any optimization: ask LLMs what they say about you. Your prices, your positioning, your offers. You will almost always find at least one inaccuracy that costs you prospects, and it is often the most profitable AEO task of the first weeks.
We activated the tool's beta on our own portal as soon as it launched. Better to test it on ourselves before recommending it to our clients. The first finding was uncomfortable: around 80% of our citations in LLM answers came from third-party sources, first and foremost the HubSpot partner directory (where Markentive has the largest number of reviews in Europe with the best rating, which explains this presence).
Good news on the face of it. Except it is not. If LLMs stop treating this directory as a reference source tomorrow, this visibility disappears, and there is nothing we can do about it. Our position on this is firm: if more than half of your citations depend on a third-party platform, you do not have an AEO strategy, you have a dependency. The task was therefore to develop citations on our own media.
The plan we deployed, which can be applied to any domain:
Result: in two to three weeks, citations obtained on our own media doubled, and our domain became more cited than our competitors' on the targeted prompts in owned media. Seven pieces of content reworked. Three weeks. A doubling measured daily by the tool, not estimated.
An honest limitation of this experience: we are talking about an already established domain, with a base of existing content to optimize. A recent site without editorial depth will not double its citations in three weeks; it will first have to build the material.
Three recommendations from the webinar, in the order in which we apply them in our projects.
First, check what the engines say about you. Hallucinations about prices, confused offers, outdated positioning: these issues can be corrected quickly (see the HubSpot pricing case) and the gain is immediate.
Prioritize depth and relevance. Specific content by target, industry and buyer journey stage, rather than generic pages. This is the direct consequence of the 350-word prompt: the LLM can match a very precise intent, as long as some content answers it.
Structure content to be scanned. Question-based headings, built-in FAQs, lists, tables, and above all unique proprietary statistics: LLMs readily cite a figure that only you have, and attribute it to your brand. Every company has these figures in its CRM or client reports. Few publish them.
A final safeguard, stated plainly during the webinar: if everyone adds FAQs and proprietary data tomorrow, on-site content will no longer make the difference. Off-site authority (brand awareness, social media, influence, editorial relationships with “peers” in your industry) remains a citation factor, and will become even more of one as best practices become commonplace. AEO does not replace brand work. It makes it measurable, prompt by prompt.