How HubSpot became the #1 CRM in AI search: what the case actually measures

HubSpot's case study reports that its Marketing team started working with the AEO tool XFunnel in June 2025 and, by the end of that year, counted a 433% increase in citations, an 1,850% increase in qualified leads from AI, and leads from AEO converting at 3x the rate of other sources. What matters for anyone weighing a geo agency in singapore is less the headline than the way each figure was defined and counted, which the post lays out closely enough to reuse as a buying checklist.

The post was published on 11 November 2026 and written by Rory Hope. It describes an eighteen-month programme rather than a one-off campaign, and its sequence is the part worth borrowing: define the buyer's journey, pick the prompts, set up measurement, then only change content. Read the numbers against those definitions before reading them as outcomes.

• AI visibility, share of voice, citation count and citation share are four separate metrics, and HubSpot reports them separately.
• On-site content was changed first, because the data showed visibility was already strong while citations were weak.
• Off-site work and forum activity came after the measurement found who was already being cited.
• The lead figures are the brand's own reporting, not an audit, and the post does not attribute them to a controlled test.

Table of contents

• What the case measures, in the order it measures it
• On-site changes and the results the post gives
• Off-site and forum work in two markets
• Reading the lead numbers with a supplier
• When the answer engines were sampled
• Is SEO dead or evolving in 2026?
• What Topkee's GEO service looks like next to this case
• FAQ

What the case measures, in the order it measures it

The first question the post says HubSpot had to answer was whether it appeared in an answer at all, so the team mapped the buyer's journey onto prompts: awareness prompts about a problem, consideration prompts that compare vendors, evaluation prompts asking what users say on review sites, and decision prompts about a specific feature. Each prompt was tracked against a defined stage, which matters because a brand can look strong at awareness and invisible at decision without noticing.

The measurement was then split into a brand-level container and a set of product containers, with feature views inside the product ones, so that a product team could run its own experiments while the wider team kept a single view. HubSpot describes four KPIs at that level: answer engine visibility, answer engine share of voice, answer engine citations, and answer engine citation share. Visibility and citations answer different questions, and the case turns on the gap between them.

A supplier that reports only the first row is reporting awareness, not influence. The post is explicit that being cited raises the likelihood of shaping an answer and of drawing direct traffic from AI; it does not claim a citation guarantees either. Any proposal from a geo agency in singapore should therefore separate the rows before it merges them into one score.

On-site changes and the results the post gives

The diagnosis was that AI visibility scores were strong from the start while citation scores were weak: the engines knew the brand but were not often quoting its pages. The work that followed aimed at specificity rather than volume. HubSpot's Growth team concluded that the answers generated for buying questions were more granular than the pages available to support them.

Three moves are described. Industry solutions pages were produced at scale from an existing library of case studies, generated by a content system and reviewed by people before publication, with breadcrumb and FAQ structured data added. Software comparison articles followed for target industries. An FAQ glossary covered top-of-funnel terms, with each page carrying a short definition, related questions and links to features.

The figures attached to each move are distinct, which is the useful part:

The directions differ and that is the lesson. The industry pages moved visibility more than citations; the comparison articles moved citations far more than mentions. A single blended score would have hidden both effects and made it impossible to decide which kind of page to fund next - the reason the case reads best as a specification for what a geo agency singapore buyer should be shown each month.

Off-site and forum work in two markets

The benchmark data showed third-party pages shaping answers about HubSpot's products. HubSpot's partnership team used the same tooling to find publishers already winning citations without naming the brand, then gave those partners templates and recommendations for answer-friendly content. By the end of 2025 the post reports partnerships with hundreds of websites, nearly a thousand new pages and hundreds of thousands of new AI citations that mention HubSpot. Those counts carry no percentage and no baseline, so treat them as volume rather than as a defensible rate.

Forum work was more measurable. Reddit came up as one of the most cited sources for the tracked prompts, so the team monitored citations weekly, identified high-impact subreddits, noted where the brand was missing and asked community advocates to address top buyer questions. In two markets with strong Reddit citation growth and no brand mentions, localised campaigns ran: within one month, HubSpot's mention rate moved from nothing, to 33.5% in France and 17.1% in Germany, and Reddit-driven citations grew from 178 in May 2025 to 146K in December 2025.

The method here is what a buyer should copy from a supplier: the gap was named per market, the intervention ran for a stated period, and the result was reported as a mention rate, not an impression count. A geo agency in singapore that reports only a global mention figure leaves the reader unable to tell whether one market or all of them moved.

Reading the lead numbers with a supplier

HubSpot reports an 1,850% increase in qualified leads from AI and a 3x conversion rate for those leads against other sources. Both are large enough to be worth querying, and the post does not describe a control group, a holdout period or an independent audit. The same page reports a 433% increase in citations, which is the aggregate behind the per-format figures above.

Market context suggests why the channel is being funded. HubSpot's State of AEO in 2026 report, drawn from 4K+ global marketers and citation themes from 6+ answer engines, finds that 58% of marketers say their businesses are optimising content for answer engines and that 44% of marketers have made a business purchase based on brands they first discovered through answer engines. Those are survey answers from a self-selected sample, not measured behaviour, but they explain the budget conversation a marketing lead will have in Singapore.

Four questions are worth putting to a supplier before accepting a lead figure, and a fifth sits behind them: whether the leads are traced to closed revenue at all, which is what Topkee's revenue tracking approach sets out to connect. The first four:

1. Which answer engines were sampled, and was each counted separately?
2. Was the prompt set fixed for the whole period, and does it include question-led prompts without the brand name inserted?
3. Is the lead number from the supplier's dashboard, the client's CRM, or a blended report?
4. What was the baseline period, and what else changed in it?

When the answer engines were sampled

A citation rate is only comparable if the sampling conditions are fixed. Two runs against the same engine can return different sets of sources, so a benchmark should state the question set, the engine, the language, the region and the cadence at which answers were collected. HubSpot's post names the engines it tracked and the buyer stages it mapped, but not a sampling cadence, and it does not publish the prompt list.

This is where a reader can apply pressure. Ask for the prompt list before the contract, not after the first report. Ask whether brand-led prompts and question-led prompts are reported apart, because adding the brand name to a question inflates any mention rate by construction. Ask whether the citation figure counts pages retrieved by the engine or pages visibly cited in the answer; the two are different and only the second is what a reader of the answer sees.

A reasonable deliverable set for a quarterly review, based on the case's structure: visibility and citation counts per engine, competitor comparison on identical prompts, a list of pages that were cited and pages that were not, and the content changes made against each gap. Ask anyone proposing a geo agency singapore retainer for the first four from a past engagement before signing.

Is SEO dead or evolving in 2026?

Search engine optimisation is not dead, and it is not sufficient on its own. The HubSpot case separates the two measurements rather than replacing one with the other: visibility and citations in answer engines are counted as their own KPIs, and the pages that earned the citations are ordinary site content: industry pages, comparison articles, a glossary and product pages. That is the argument in [AEO vs. Search Rankings: AI Visibility Basics](https://www.topkee.com.sg/dynamic/ai-visibility-vs-search-rankings-is-your-agency-measuring): different reports on the same pages.

Look at where the reported lifts came from. The industry pages, the comparison articles and the product pages are all pages on HubSpot's own domain, edited for structure, tables, FAQs and structured data. That is the same work a conventional SEO programme does, aimed at a different consumer of the page. Meanwhile the citations metric tracks whether those pages are actually referenced, which a rankings report cannot show even when a page ranks first.

The practical reading for a Singapore marketing lead is that SEO and AI-search work share a base and diverge at the report. A site that is poorly structured, slow or vague will not be cited, and a site that ranks well can still go uncited. The case supports the second statement directly: HubSpot's visibility scores were strong from the jump while its citation scores were weak. Nothing in the post suggests that organic rankings stopped producing traffic, and nothing in it shows a ranking result standing in for a citation result.

What changes in 2026 is where the audit happens. An SEO report asks which queries brought impressions and clicks. An AI-visibility report asks which prompts the brand appears in, who else appears, and which pages the engine cited. Both are needed, and a supplier that offers only one should say so.

What Topkee's GEO service looks like next to this case

Topkee runs a GEO service built around the same measurement-then-content order, scoped to one brand, one primary website, one market and one content language. The work is a loop of monitoring, analysis and content delivery rather than a single content push: it tracks how the brand is mentioned in AI answers, which sources the engines draw on, and then closes the gaps with human-edited knowledge content.

The metric definitions are where the service speaks directly to the HubSpot case. Topkee defines AI visibility mention rate as valid answers containing a brand mention divided by valid answers under the same conditions, with own-site source rate calculated separately, and it holds that comparisons require a fixed question set, engine, language, region and sampling cadence, with question-led and brand-led samples kept apart. Its GEO Insight console shows brand ranking, visibility, competitors and citation sources inside AI answers, with active-prompt quotas that rise across three monthly plans, and monthly knowledge pieces in the same stepped pattern.

Three limits are stated plainly and are worth carrying into any conversation with a supplier. Engines are sampled through an API, so the results are not identical to what a consumer sees in the retail product. Citation summaries include non-inline retrieval sources, so they should not be presented as a visible citation rate. And the sample report in the knowledge base is simulated, not a client result, so it cannot be used as evidence of an outcome. Off-site media placement and CMS publishing sit outside the standard monthly scope, which is a meaningful difference from the second and third pillars of the HubSpot case. Topkee publishes plans for the Hong Kong market; its Singapore office is a company presence, and no Singapore-specific plan is recorded.

For a Singapore team, that boundary is the first thing to test in a proposal, and the leads question has its own answer path - revenue tracking that joins enquiries to closed value - before a citation count is read as a business result. A geo agency singapore offer that promises third-party placements and forum activity should say whether those are custom scope, and which of the four KPIs it will report each month.

FAQ

Is the HubSpot case an independent study?

No. It is the brand's own account, written by Rory Hope and published on the HubSpot Blog on 11 November 2026. The measurements are described in enough detail to be checked against a supplier's method, but the lead and citation figures are self-reported and no control group is described.

What is the difference between AI visibility and citations?

Visibility counts whether the brand appears in an answer for a target prompt; citations count whether its pages are quoted as a source. HubSpot's data showed strong visibility with weak citations, which is why the content programme targeted specific, citable pages next.

How long did the programme take to show results?

The team started with XFunnel in June 2025. The France and Germany mention rates moved within one month of the localised campaigns, and the Reddit citation count is given from May 2025 to December 2025, so plan for months rather than weeks.

Do I need a separate agency for GEO and SEO?

Not necessarily, but the report should be separate. The case measures rankings and citations as different things on the same pages, so ask any supplier to show which metric each change was expected to move and what the baseline was.

What should I ask a GEO agency for in the first meeting?

The prompt list, the engine list, whether brand-led and question-led prompts are reported apart, how mention rate and own-site source rate are calculated, and which deliverables are inside the monthly fee. The one-year customer figures on product pages are not a substitute for that method.

Sources

1. Rory Hope, "How HubSpot became the #1 CRM in AI search [A case study]", HubSpot Blog, 2026-11-11. https://blog.hubspot.com/marketing/hubspot-aeo-case-study
2. "State of AEO in 2026", HubSpot, undated. https://offers.hubspot.com/state-of-aeo