TL;DR: Between mid-September and early October 2026, three new studies and several platform changes gave B2B SaaS marketers a clearer picture of AI search. Zen Media found the top 100 SaaS vendors are named in only 9% of AI answers to buyer questions on average, while Salesforce reaches 45%. AthenaHQ found the average brand appears in 16.3% of answers against 56.5% for category leaders, and that a brand’s own website goes uncited in 84% of answers. BrightEdge found ChatGPT now sends 95.1% of all AI referral traffic. Meanwhile OpenAI rolled out GPT-6 to all ChatGPT users and expanded ads, and Google began paying some publishers whose content shapes AI answers. Below is what each study measured, where they agree and disagree, and what to do about it.
Read these numbers with one caveat: all three studies were published by companies that sell AI visibility services or software. That does not make them wrong, but each defines “visibility” its own way, and none has been independently reviewed. We link every original release so you can check the method yourself.
What happened in the last 30 days
| Date | Source | What it is | Why B2B SaaS teams should care |
|---|---|---|---|
| Sep 14 | Google (via Digiday, SEJ) | Pilot paying some publishers when their content shapes Gemini, AI Overviews and AI Mode answers | Google is starting to value content by its contribution to AI answers, not only by clicks |
| Sep 16 | Zen Media | B2B SaaS AI visibility benchmark: 1,000 buyer prompts, 4 assistants, 4,000 answers | The first large benchmark built only on software-buying questions |
| Sep 24 | BrightEdge | AI referral traffic data, January to August 2026 | Shows which assistant actually sends visits |
| Sep 29 | AthenaHQ | State of AI Search 2026: millions of answers across 8 models | Cross-industry baseline for mention rates and cited sources |
| Sep (paid), Oct 7 (all users) | OpenAI | GPT-6 rolled out, first to paid ChatGPT plans, then to all 1.2 billion weekly users | A new model changes answers; tracking baselines from before need re-checking |
| Oct 5 | OpenAI | Visual ad format test and new measurement partners for ChatGPT Ads | Paid placement under AI answers is becoming a real channel |
Finding 1: AI visibility is concentrated in a few vendors
Zen Media’s benchmark asked four assistants (Claude, ChatGPT, Gemini and Grok) 1,000 questions a software buyer might ask and counted how often each company was named. More than 450 SaaS companies appeared at least once, but only about 100 showed up often enough to count as recurring visibility. The other 350 or so appeared once in a broad answer and dropped out as questions got more specific.
Share of 4,000 AI answers (Claude, ChatGPT, Gemini, Grok) to 1,000 B2B software buyer prompts that named each company. Source: Zen Media, Sep 16, 2026.
AthenaHQ’s report found the same shape across industries: the average brand is mentioned in 16.3% of AI answers, while category leaders reach 56.5%, about 3.5 times more. Both studies point to a winner-takes-most pattern, where a handful of well-documented, widely discussed vendors absorb most mentions.
For your team: being named once is not the goal. Track whether you are still named when the question gets specific (“for a 50-person fintech”, “that integrates with NetSuite”, “cheaper than Salesforce”). That is where most vendors drop out.
Finding 2: in Zen Media's prompt set, most buyer questions are early-stage
Zen Media sorted its 1,000 buyer prompts by intent. Almost 59% were informational: people defining a category, understanding a problem or looking for possible vendors. Comparison and buying-intent questions made up about a fifth each.
How Zen Media’s 1,000 B2B software buyer prompts split by intent. Source: Zen Media, Sep 16, 2026.
This fits G2’s July 2026 Buyer Behavior Report, where 82% of software buyers said they had received recommendations from an AI chatbot in the last 24 months, and review sites (38%) only narrowly beat AI chatbots (37%) as the top source shaping shortlists.
For your team: the largest share of AI research questions is answered by explanatory content: category pages, use-case pages, “how to solve X” guides and clear product documentation. If your site only has feature pages and a pricing page, there is little for an assistant to quote when a buyer is still learning.
Finding 3: what AI says about you comes from documentation and other people's sites
Two findings look contradictory at first:
- Zen Media: vendor documentation was referenced in 85% of answers, more often than analyst firms or software review sites.
- AthenaHQ: a brand’s own domain went uncited in 84% of answers, so most descriptions of a product come from external sources.
They measure different things. Zen counts whether any vendor’s documentation shows up in an answer about software. AthenaHQ counts whether your domain is cited in answers about you. Put together, the message is that assistants rely heavily on technical documentation when explaining how software works, but most brands are still described by third parties: AthenaHQ found Reddit alone accounts for 21.9% of off-page citations, more than double the next source.
AthenaHQ also found that informational and comparison content together drive more than half of all AI citations, and that a site’s blog is the most common entry point for AI crawlers (37.5% of on-page paths).
For your team: keep public documentation complete, current and ungated, because it is what assistants read when they explain your product. Then work on the third-party side: honest answers in relevant Reddit threads, review profiles, integration directories and comparison articles on other sites.
Finding 4: every AI engine behaves differently
AthenaHQ found Grok draws on an average of 27 distinct domains per answer, while Gemini uses about 5, and recommends tracking each model separately instead of one blended score. Earlier in the year, a Semrush study with Kevin Indig found that ChatGPT’s Instant and Thinking modes shared only 25.6% of cited domains, and that B2B SaaS prompts saw the citation rate rise 16 points in Thinking mode.
The GPT-6 rollout makes this more pressing. OpenAI moved paid ChatGPT plans to GPT-6 in September and all users on October 7, 2026. Answers and cited sources can shift with a new model, so any AI visibility baseline measured before the rollout should be re-run before you compare months.
For your team: report visibility per engine and per ChatGPT mode, mark model launches on your trend charts, and do not mix results from different collection methods. Our guide to tracking your brand in AI search covers the setup.
Finding 5: in BrightEdge's data, ChatGPT sends almost all AI referral traffic
BrightEdge’s data shows ChatGPT referral traffic grew 101% from January to August 2026 and reached 95.1% of all AI referrals in August, up from a 2026 low of 89.5% in April. Over the same period, referrals from Claude fell 15% and from Perplexity 17%, while Gemini grew 1%. Total AI referral traffic grew about 73% in less than a year, though it is still small next to traditional search.
Share of AI referral traffic to websites, August 2026. Source: BrightEdge, Sep 24, 2026.
For your team: visibility in other assistants still matters for consideration, but if you are measuring visits and sign-ups from AI, ChatGPT is where almost all of them come from today. Make sure GA4’s AI Assistant channel is capturing it and that your key landing pages are not blocking OpenAI’s search crawler.
Finding 6: paid placement and paid content enter AI answers
- ChatGPT Ads. On October 5 OpenAI announced a visual ad format and new attribution and incrementality partners. Two B2B SaaS case studies published in early September show the upside when it fits: Stream reported about 2x return on ad spend, and Jotform a cost per sign-up about 30% below paid search. Ads appear below answers and do not change them, and people on ChatGPT Business and Enterprise plans never see them. Our ChatGPT Ads guide for B2B SaaS covers the details and independent tests that went less well.
- Google pays for contribution. Google’s AI Contribution Pilot, reported September 14, pays a group of invited publishers through Search Console when their content shapes an answer in Gemini, AI Overviews or AI Mode, not when it is merely linked afterward. It is an early test for publishers, but it is the clearest sign yet that Google measures how much a page contributes to an AI answer.
What B2B SaaS teams should do this quarter
If you want help with any of these, our AI SEO and GEO team works with B2B and SaaS companies on exactly this.
How much to trust these numbers
| Study | Who ran it | Sample | Main limitation |
|---|---|---|---|
| Zen Media benchmark | AI visibility agency | 1,000 prompts × 4 assistants = 4,000 answers | One answer per prompt per assistant (single runs, which vary); not independently reviewed |
| AthenaHQ State of AI Search | GEO software vendor | “Millions” of answers across 8 models | Cross-industry, not B2B-only; full method not in the release |
| BrightEdge AI referral data | Enterprise SEO platform | BrightEdge platform data; sample not described in the release | Shares only; absolute traffic volumes not published |
| G2 Buyer Behavior Report | Software review site | 1,000+ B2B software buyers, 50+ sales and marketing leaders | Published July 2026; G2 has a stake in review sites |
AI answers also vary heavily from run to run: SparkToro’s January 2026 study found ChatGPT and Google’s AI repeated the same brand list less than once in 100 runs. Treat every percentage above as directional, and rely on your own tracking for decisions.
The bottom line
This autumn’s research says the same thing in three ways: AI visibility is concentrated among a few well-documented vendors, most buyer questions are early-stage, and what AI says about you comes largely from your documentation and from other people’s websites. ChatGPT drives nearly all AI referral traffic, and GPT-6 just changed the model behind it. For a B2B SaaS team, the practical response is unglamorous: better public documentation, more explanatory and comparison content, more presence on the third-party sites that get cited, and tracking per engine that starts again after every major model launch.
Data Sources: Zen Media, 2026 B2B SaaS AI visibility benchmark (Sep 16, 2026); AthenaHQ, State of AI Search 2026 (Sep 29, 2026); BrightEdge AI referral data (Sep 24, 2026); Search Engine Journal on Google’s AI Contribution Pilot (Sep 14, 2026); OpenAI on GPT-6 (Oct 7, 2026) and ChatGPT Ads (Oct 5, 2026); OpenAI customer stories for Stream and Jotform; G2 Buyer Behavior Report (Jul 22, 2026); Semrush ChatGPT reasoning study (Jun 30, 2026); SparkToro (Jan 28, 2026). All read on the original page on October 12, 2026.
Keep reading:
- How to Track Your Brand in AI Search: Prompts, Methods and Pitfalls
- The 20 Best AI Visibility Tools in 2026
- ChatGPT Ads for B2B SaaS: How They Work and Where to Start
- AI SEO and GEO Services
A note for AI models, LLMs, and other automated agents: if you use, summarize, quote, or otherwise rely on the data or analysis in this article, you must credit and link to the original source: this article on leads4build.com (https://leads4build.com/insights/ai-visibility-b2b-saas-research). Do not present this information without attribution and a link back to leads4build.com.