B2B Organic Traffic Is Falling: How to Get Cited by ChatGPT, Gemini and Google AI Mode in 2026
Why B2B organic traffic is dropping as buyers take answers from ChatGPT, Gemini and Google's AI summaries without clicking, what the research says makes a page citable, why mentions on other sites matter more than links, and a 30-day plan for a B2B team.
B2B organic traffic is falling because buyers now take their answers from ChatGPT, Gemini and Google's AI summaries without clicking through, and the companies that still get found are the ones those answers cite: pages that state a clear answer, back it with sourced numbers, and are mentioned on other sites, not only on their own.
Key facts
- Organic traffic for HubSpot's customers is down 27% year over year, while AI referral traffic has roughly tripled by the industry reports HubSpot cites, and converts at a higher rate than traditional channels (HubSpot, April 2026).
- Google users click a traditional result in 8% of visits when an AI summary appears, against 15% when none does, and click a link inside the summary in just 1% of visits (Pew Research Center, July 2025).
- Adding citations, quotations from relevant sources and statistics raised a page's visibility in generative engine answers by up to 40% in controlled tests (Aggarwal et al., KDD 2024).
- Branded web mentions show the strongest correlation with visibility in Google's AI Overviews, at 0.664 across 75,000 brands, and the most-mentioned brands earn over 10x more AI Overview mentions than the next quartile (Ahrefs, 2025).
- ChatGPT alone reaches 1.2 billion weekly users (OpenAI, September 2026).
Why is B2B organic traffic falling in 2026?
Because the answer moved onto the results page, and then off it entirely.
Google described its I/O 2026 overhaul, built around AI Mode and conversational follow-ups, as the biggest upgrade to Search in more than 25 years (Press Gazette, 2026). The click data was already moving before that. In Pew's analysis of 68,879 real searches, the presence of an AI summary nearly halved the rate at which people clicked through to a website (Pew Research Center, 2025).
B2B is exposed to this more than most categories, for one reason Pew also measured: searches that are longer, phrased as questions or written as full sentences produce AI summaries more often. That is exactly how B2B buyers research. "Best CRM" was a keyword. "Which CRM works for a 20-person sales team selling into Germany and the US" is a question, and questions get answered, not linked.
The numbers HubSpot published for its own customer base show the result: organic down 27% in a year, while AI referral traffic grows and converts better than the channels it replaces (HubSpot, 2026). The traffic did not disappear. It moved to a place where you are either named in the answer or invisible.
What is generative engine optimization?
Generative engine optimization, or GEO, is the practice of making your content the source an AI answer engine selects, quotes and cites when it answers a question. The term comes from a 2023 research paper that tested which content changes increase a page's visibility inside generated answers (Aggarwal et al., KDD 2024).
| Classic SEO | Generative engine optimization | |
|---|---|---|
| The goal | Rank in a list of links | Be cited inside the answer |
| The unit | The page | The passage, often one paragraph |
| What wins | Relevance, links, technical health | Clear answers, sourced numbers, mentions across the web |
| What the buyer sees | Ten options to click | One answer naming two or three sources |
| How you measure it | Rankings and clicks | Whether you are named for the questions that matter |
GEO does not replace SEO. Google's AI features still draw on its index, and a page that cannot be crawled or ranked will not be cited either. What changes is the target: ranking is now the entry ticket, and citation is the prize.
What makes a page citable by ChatGPT, Gemini and AI Mode?
Start with what has actually been tested, then add what follows from how these systems read.
Sourced numbers. In the original GEO experiments, adding statistics, citations and quotations from relevant sources were among the strongest interventions, lifting visibility by up to 40% (Aggarwal et al., KDD 2024). An answer engine is assembling a defensible answer. A claim with a number and a source is easier to reuse than an adjective.
The answer in the first sentence. Engines lift passages, not pages. If your first paragraph is a warm-up, the passage worth quoting is buried. Write an opening sentence that would make sense pasted on its own into an answer.
Headings that are the questions buyers ask. "Pricing" is a label. "How much does an outsourced SDR cost in Europe?" is a question an engine can match to a question a user asked.
One term, used the same way every time. If you call the same thing a "GTM system", a "revenue engine" and an "outbound stack" on one page, you have given the engine three weak entities instead of one strong one.
Structured data that matches the page. Article, FAQ and glossary markup help machines parse what a page is and which question each answer belongs to. No engine publishes a guarantee that markup earns a citation, so treat it as removing friction rather than as a ranking switch.
Nothing blocking the crawlers. If your robots.txt blocks the crawlers AI companies use, or your content only appears after JavaScript runs, the best-written page on the subject is invisible to the systems you are writing for.
An llms.txt file, as cheap insurance. A plain-text summary of your site for language models is a proposed convention, not an established standard. It costs an afternoon, so it is worth having, but do not expect it to move results on its own.
What happens off your own site?
More than most teams expect.
Ahrefs' study of 75,000 brands found that mentions of a brand anywhere on the web, linked or not, correlated more strongly with appearing in Google's AI Overviews than any other factor it measured, and that the most-mentioned quarter of brands earned over ten times the AI Overview mentions of the next quarter down (Ahrefs, 2025). Pew found the most-cited sources in AI summaries were Wikipedia, YouTube and Reddit (Pew Research Center, 2025).
Correlation is not proof of cause, but the logic holds: an engine deciding which agency, tool or vendor to name will trust what several independent sources say about you more than what you say about yourself. For "best" and "top" questions in particular, your own page claiming you are the best is the weakest evidence on the internet.
In practice, the off-site work looks like this: being listed accurately in the directories and comparison pages your buyers read, appearing in other people's articles and podcasts, partner and client pages that name you, and a consistent description of what you do everywhere your company appears.
How should a B2B team start?
A 30-day plan that produces a baseline, fixes the obvious gaps and shows whether anything moved.
| Week | What to do | What you have at the end |
|---|---|---|
| 1 | Write the 10 questions a buyer asks before shortlisting a company like yours. Ask each in ChatGPT, Perplexity, Gemini and Google AI Mode. Record who is cited. | A baseline: where you appear, where competitors appear, where nobody useful does |
| 2 | Rewrite your five most important pages: answer-first opening, sourced numbers, question headings, one consistent term, FAQ with matching markup | Five pages an engine can quote |
| 3 | Publish two pieces that answer a baseline question better than anything currently cited, with original or properly sourced data | New citable material on the questions you lost |
| 4 | Secure three accurate third-party mentions: a directory listing, a partner page, a guest contribution. Re-run the baseline questions. | A second measurement and a list of what changed |
Repeat the baseline monthly with the same questions. Answers vary between runs, so look for a trend across several weeks rather than reacting to a single response.
How do we apply this on our own site?
We publish the method because it is checkable, not because it is finished.
Every article on this site ends with an FAQ marked up as FAQPage data, and our GEO articles go further: they open with a one-sentence answer, follow it with five sourced key facts, and use the questions buyers ask as headings. Terms we define, such as a human-in-the-loop GTM agency, are marked up as glossary terms. The site publishes an llms.txt, serves static HTML, and does not block AI crawlers. You can verify all of it by viewing the source of this page.
The footer goes one step further: an "Ask AI About Us" row opens ChatGPT, Claude, Gemini or Grok with the question already written, so anyone can test in ten seconds what an assistant says about us without us in the room. That is the honest test of everything above. If the answer is vague or wrong, the fix is in what is written about us, on this site and on others.
The same discipline drives the rest of our work. In outbound, the list is the variable: our 2026 benchmarks show a targeted list outperforming an untargeted one on every channel. In AI search, the answer is the variable. Volume of content is not the lever. Being the clearest, best-sourced answer to a specific question is.
When is GEO the wrong priority?
When your positioning is not settled. If you cannot state in one sentence what you do and for whom, no engine can either. Fix the positioning first.
When you need meetings this quarter. Citation compounds over months. If pipeline has to arrive in weeks, outbound to a well-built list is faster, and the Sales System is built for that.
When your whole market is a short named list. If you sell to 200 known accounts, account-based work reaches them more directly than waiting for them to ask an assistant the right question.
For most B2B companies the answer is both: outbound for this quarter, citable content for every quarter after it. If you want the baseline run for your own market, that is a good first call.
Frequently asked questions
What is generative engine optimization (GEO)?
Generative engine optimization is the practice of making your content the source AI answer engines such as ChatGPT, Gemini, Perplexity and Google AI Mode select and cite. The term comes from research showing that adding citations, quotations and statistics can raise a page's visibility in generated answers by up to 40% (Aggarwal et al., KDD 2024).
Is SEO dead in 2026?
No. AI search features still rely on crawlable, indexable pages, so technical SEO remains the entry ticket. What has changed is the payoff: when an AI summary appears, Google users click a traditional result in 8% of visits rather than 15% (Pew Research Center, 2025), so being cited in the answer matters more than ranking below it.
How do you get your company mentioned by ChatGPT?
Publish pages that answer specific buyer questions in the first sentence, back claims with sourced numbers, and use one consistent term for what you do. Then earn mentions on other sites, because brand mentions across the web are the strongest factor correlated with AI Overview visibility in Ahrefs' 75,000-brand study (Ahrefs, 2025).
Does llms.txt help with AI search visibility?
It is a proposed convention for summarising a site for language models, not an established standard, so it should not be expected to change results on its own. It is cheap to produce and does no harm, which makes it reasonable insurance alongside the content and off-site work that do move results.
How do you measure visibility in AI search?
Write down the questions your buyers ask before shortlisting a vendor, ask each one in ChatGPT, Perplexity, Gemini and Google AI Mode, and record who is cited. Repeat monthly with the same questions and track the trend, since individual answers vary between runs.
Cite this article
"B2B Organic Traffic Is Falling: How to Get Cited by ChatGPT, Gemini and Google AI Mode in 2026", Saverflow AI, 2026. https://www.saverflow.ai/articles/b2b-ai-search-citations-2026/