Foundations
They run HubSpot and buy Google Ads. AI still does not name them.
By Arnav Mukherjee, founder of TofuBofu · August 5, 2026
Earlier this month we pulled the IT services firms our regional research had already measured as missing from AI vendor answers. We enriched 72 of them, looking for the ones worth a conversation. I thought I knew what the data would say. Companies that never got round to marketing: no CRM, a dated website, a founder doing sales between support tickets. That is the story the category tells about itself, and it is the story I half believed.
The data said something else. Of the 44 firms where any marketing technology was detected at all, 27 were running HubSpot marketing automation and 27 were running HubSpot CRM. Twenty-eight ran their public website on HubSpot CMS. Several were buying Google Ads on top. One 28-person firm in Irvine had the full HubSpot suite plus roughly 672 dollars a month of detected ad spend. A 36-person firm in Florida had the same suite, plus roughly 1,479 dollars a month.
These are not firms that forgot to do marketing. They are firms that bought the recommended stack, staffed it, and are still absent from the answer their buyers now get first. That gap is worth understanding properly, because the obvious explanations are the wrong ones.
What the 72 firms were actually running
The enrichment came from a third-party data provider. Treat the detections as good signals rather than audited fact, and the ad figures as estimates rather than anyone's confirmed budget. With that caveat, the distribution is striking.
The first row is the most useful one. Twenty-eight of these firms run their website on HubSpot CMS. That means clean templates, fast pages, sane markup and complete control over what a crawler sees. So whatever is keeping them out of AI answers, it is mostly not that the engines cannot read their site. The comfortable technical explanation is gone, and a harder question takes its place.
The stack works below the line where the shortlist gets made
Marketing automation is demand capture. Every capability it sells operates on a person who has already arrived: forms, lifecycle stages, lead scoring, nurture sequences, attribution reporting. It is genuinely good at that. That is why so many of these firms bought it.
Paid search sits slightly higher up. It still buys placement against a query the buyer has already formed, and it stops the moment the card stops being charged. An AI recommendation happens earlier than either. It happens when a buyer asks who they should be considering at all, and the answer is assembled from evidence the engine retrieves about you. There is no inventory to buy. No bid, and no account manager to call.
So a firm can own every tool in the funnel and be structurally absent from the moment that decides who enters it. Worse, nothing in the stack reports this. The dashboard measures people who arrived. Buyers who never saw your name never arrive, so they never get a contact record and never appear in a single report. The number goes quietly down. Every chart says the funnel is healthy.
Where the stack operates, and where the shortlist is decided
What this data can and cannot tell you
I want to be careful here, because the number is easy to misuse and I would rather hand you the honest version. These 72 firms were selected precisely because earlier research had already found them missing or only partly present in AI answers. That is a description of the invisible group, not a rate across the industry.
So this cannot tell you what share of HubSpot customers are invisible in AI search. It emphatically does not show that owning a marketing stack causes invisibility. Firms with good stacks may well be cited more often than firms without. Nothing here measures that, because well-cited firms were never in the sample.
What it does kill is one specific, widespread assumption. That the companies missing from AI answers are the ones who never invested in marketing, and that buying the stack is therefore the fix. Within the invisible group, a clear majority of the firms with any detectable tooling had bought exactly that stack. Whatever separates a cited firm from an uncited one, it is not the licence.
Find out which side of the line you are on
Run a free scan across six AI engines on your own buying question, and see whether your firm appears in the shortlist before the funnel ever starts.
Get your free auditWhat does move the answer
If the stack is not the lever, the useful question is what is. The short answer is corroboration: several independent sources saying a consistent, specific thing about you. Four things follow from that, and none of them requires new software.
Say the specific things only you can say
Which cities you serve, which industries you are actually good at, the size of client you fit, your compliance posture, your response commitments. Most firms hold all of this internally and have never written it down in one retrievable place. An engine cannot repeat a fact it cannot find, and a page that says you deliver world-class IT solutions gives it nothing to work with.
Make the third-party record agree with itself
Directory entries, review profiles and association listings are what an engine reaches for when it needs a second opinion about you. If your name, location and category differ across them, you are harder to verify than a competitor whose listings match. Profound found brands present across four or more platforms were about 2.8 times more likely to be cited.
Treat reviews as evidence, not decoration
Kevin Indig's analysis with G2 found roughly 10 percent more reviews associated with about 2 percent more AI citations, and G2 holds about 22.4 percent share of voice for software. The mechanism is not flattery, it is that reviews create independent text about you that an engine can retrieve and quote. Real reviews only: fabricated ones violate the platforms' terms and get removed.
Measure the moment above the line separately
Your existing analytics cannot see this, by construction. Track which engines name you on your real buying questions, per engine, over time, alongside the funnel metrics you already watch. Otherwise a shortlist problem and a conversion problem produce the same shrinking number and you will spend a year optimising the wrong one.
The framing we keep coming back to is that search rank is the floor and AI visibility is a distinct layer built on top of it. These 72 firms are the clearest illustration I have seen of the gap between the two. They cleared the floor comfortably. Real websites, real budgets, real tooling, in many cases real Google rankings. The layer above it was never something the stack was going to build for them.
G2's 2026 research found 51 percent of B2B buyers now begin vendor research on an AI chatbot, up from 29 percent, and Forrester's 2026 study found 94 percent use AI somewhere in the buying process. The share of buyers who form their shortlist before touching anything a marketing stack can measure is growing. That is the whole argument, and the 72 firms are what it looks like in practice.
Frequently asked questions
Does having HubSpot improve your AI visibility?
Not on its own. We enriched 72 IT services firms that our regional research had already measured as missing or only partly present in AI vendor answers. Of the 44 where marketing technology was detected, 27 were running HubSpot marketing automation and 27 were running HubSpot CRM. Owning the automation platform did not correlate with being named, because the platform operates on people who have already found you. It nurtures, scores and routes existing contacts. It does not create the third-party evidence an engine needs before it will say your name.
Why does paid search not help with AI visibility?
Because they buy different moments. Paid search buys placement against a query a buyer has already formed, and stops the moment you stop paying. An AI recommendation happens earlier, when the buyer is asking who they should even consider, and it is assembled from evidence the engine retrieves rather than inventory anyone can buy. There is no auction. In our sample, two firms with detected monthly ad spend of roughly 672 and 1,479 US dollars were still absent from the answers to their own regional buying question.
If my website is on a modern CMS, why am I not cited?
Because a good CMS solves the publishing problem, not the corroboration problem. In this sample 28 of the firms ran their website on HubSpot CMS, meaning clean templates, fast pages, working markup and total control over the crawlable surface, and they were still missing from the answers. That is genuinely useful information: for most of these firms the bottleneck was not technical access. It was that nothing outside their own domain said anything specific enough about them for an engine to repeat.
What actually makes an AI engine name a company?
Corroboration, meaning several independent sources that say a consistent, specific thing about you. Directory listings that agree on your name and location, review profiles, third-party roundups, and your own pages stating plainly what you do, for whom and where. Profound's research found brands appearing across four or more platforms were about 2.8 times more likely to be cited, and Kevin Indig's analysis with G2 found roughly 10 percent more reviews associated with about 2 percent more AI citations. None of that is something a marketing automation licence can produce for you.
Does this mean marketing automation is a waste of money?
No, and it would be a bad conclusion to draw from this data. These tools do the job they are sold for, which is converting and retaining demand that already exists. The finding is narrower: that job sits entirely downstream of the moment an AI engine assembles a shortlist. A firm can run an excellent nurture programme on a steadily shrinking pool of inbound enquiries and see nothing in the dashboard that explains why the pool is shrinking, because the buyers who never made the shortlist never arrive to be tracked.
Can this data prove that a marketing stack causes AI invisibility?
No, and we should be explicit about that. These 72 firms were selected precisely because our earlier research had already found them missing from AI answers, so this is a description of the invisible group, not a rate across all IT firms. It cannot tell you what share of HubSpot users are invisible, and it certainly does not show that the stack causes the invisibility. What it does rule out is the comfortable assumption that firms missing from AI answers are simply the ones not doing any marketing.
How do I tell if my own firm has this problem?
Ask the buying question your customers ask, in the market you serve, across several engines, and read who gets named. If your CRM shows healthy conversion on the leads you get while the number of new enquiries drifts down, that pattern fits a shortlist problem rather than a funnel problem. The two failure modes look identical in a marketing dashboard, because the dashboard can only measure people who arrived, which is exactly the population that excludes everyone the engine never told about you.
Sources and further reading
- TofuBofu Research: the regional multi-engine studies that identified these firms as absent from AI vendor answers before any enrichment was run.
- G2 2026 B2B buyer research: 51 percent begin vendor research on an AI chatbot, up from 29 percent, and 69 percent switched vendor based on AI.
- Forrester B2B Buying Study 2026: 94 percent of buyers use AI somewhere in the buying process.
- Profound: brands appearing across four or more platforms are about 2.8 times more likely to be cited.
- Kevin Indig with G2: G2 holds about 22.4 percent share of voice for software, and roughly 10 percent more reviews is associated with about 2 percent more AI citations.