Most B2B keyword lists we get sent have already been pruned by search volume before anyone looked at who is typing the query. The terms that survive are the big informational ones, and those are exactly the queries Google now answers itself.
That is the whole problem with B2B SEO in 2026 in two sentences. The rest of this post is what to do about it.
Why B2B keyword volumes are tiny, and why that is fine
A B2B purchase is made by a group, slowly, and mostly without you in the room. 6sense's 2025 Buyer Experience Report, based on more than 4,000 buyers, puts the average cycle at around ten months, with buyers doing roughly 60% of the journey before they contact a seller. 94% of buying groups had a preferred vendor before first contact, and 77% bought from that vendor (6sense).
Read that against a keyword tool. A query typed by one procurement lead, once, on behalf of a deal worth six figures, shows up as a rounding error next to a consumer query with no budget behind it.
So the volume column is the wrong filter. The filter we use is: would a member of a buying committee type this while deciding something, and what would they need to see on the page to move the decision? That produces a list of small, specific queries, most of which a tool will mark as low or zero volume, and that list is the plan.
Volume still has a job: prioritising within that list, and flagging the head terms AI answers will absorb anyway. Our keyword research tool returns volume, difficulty and CPC for a pasted list, and CPC is often the more useful column in B2B. If competitors pay real money for a term, someone is buying on it.
What AI Overviews and AI Mode did to the research phase
The research phase is where B2B content used to win, and it is where the clicks went.
Pew Research tracked 900 US adults across nearly 69,000 Google searches in March 2025. When an AI summary appeared, users clicked a traditional result on 8% of visits, against 15% when it did not. They clicked a link inside the summary on 1% of visits. And 60% of question-style queries (who, what, when, why) produced a summary, which is the shape of most B2B research queries (Pew Research Center).
Ahrefs measured it from the other side, using aggregated Search Console data for 300,000 keywords. Their April 2025 study found AI Overviews cut position-one clickthrough by 34.5%. When they re-ran it with December 2025 data the gap had widened to 58% (Ahrefs).
Seer Interactive's data across 53 brands and 5.47 million queries adds a twist. Organic CTR on queries with an AI Overview fell to a floor of 1.3% in December 2025, then recovered to 2.4% by February 2026. Being cited inside the Overview delivered roughly 120% more clicks per impression than not being cited, although cited pages still trailed non-AIO queries (Seer Interactive).
And it has not stopped moving. In late August 2026 Google started dynamically expanding some AI Overviews into full AI Mode-style answers with the "Ask anything" box loaded by default, no "Show more" click required (Search Engine Roundtable).
The practical reading for a B2B marketer: informational pages are now mostly a citation game, not a traffic game. Commercial pages are still a traffic game. Plan them differently.
Query fan-out changes what "covering a topic" means
AI Mode and AI Overviews do not run your query. They run a set of queries. Google describes the technique as "breaking down your question into subtopics and issuing a multitude of queries simultaneously on your behalf" (Google).
A buyer who types "best contract management software for mid-size legal teams" is, behind the scenes, generating sub-queries about pricing models, integrations with document systems, security certifications, implementation time and named alternatives. Your page is retrieved against those sub-queries, not against the headline phrase.
This is why the old B2B pattern of one thin page per keyword fails in AI surfaces. The page that gets pulled in is the one with the clearest answer to one fan-out question, and the site that gets pulled in repeatedly has a clear answer to most of them somewhere.
Two things follow. First, map the sub-questions for every commercial topic you care about and give each one a home on the site, as its own page or a clearly headed section. Second, do not shred content into fragments to game retrieval. Google's own AI optimisation guide says there is no requirement to "break your content into tiny pieces for AI to better understand it", and that what it wants is "unique expert or experienced takes that go beyond common knowledge" (Google Search Central). We wrote up the mechanics and what our own data showed in our query fan-out explainer.
The bottom-funnel pages that still get clicks
This is the list we build first for any B2B client, before any blog post.
| Page type | Why it still earns a click | What has to be on it |
|---|---|---|
| Pricing | Buyers want the number, and AI answers hedge on it | Real tiers or a real range, what changes the price, what is included |
| "X vs Y" | Comparison queries are late-stage and specific | Honest feature table, who each is better for, your own weaknesses |
| Alternatives to [competitor] | Someone is unhappy with an incumbent right now | Named alternatives including you, the switching cost, migration notes |
| Integrations | Technical evaluators check this before shortlisting | One page per integration, what syncs, setup steps, limitations |
| Use case / persona pages | "[Product] for [industry]" queries carry budget | Workflow specific to that industry, screenshots, a customer in that sector |
| Security and compliance | Procurement blocks deals without it | Certifications, data residency, sub-processor list, DPA link |
These pages convert because the visitor arrives with most of the decision made, which is also why they hold their clickthrough better than research content. Pew's finding that question-style queries trigger summaries most often is the flip side: "[competitor] pricing" or "[tool A] vs [tool B]" is not something Google can safely answer without sending you to the source.
If you sell software, this is most of the SaaS SEO playbook. For services businesses the same logic applies with different nouns: "vs" becomes "in-house vs outsourced", and integrations become "works with your existing accountant, lawyer or platform".
Product-led and problem-led content, not topic-led
The generic B2B blog (what is X, benefits of X, X best practices) is the content most exposed to AI answers, because it has no reason to exist beyond the keyword. We push clients toward two other shapes.
Problem-led content starts from a symptom the buyer actually has. "Our sales team keeps quoting from an outdated price book" is a problem. "Sales enablement best practices" is a keyword. The first earns a citation because it names a specific situation, and it pre-qualifies the reader.
Product-led content shows the product doing the job, with real screens, real limitations and the steps in between. AI models summarise abstract advice easily. They are a poor substitute for a walkthrough of the actual workflow, and buyers on a shortlist want that walkthrough.
Both shapes give sales something to send. In a ten-month cycle, a page that gets forwarded to the quieter members of the buying group does more work than one that ranked once for a head term.
Building for citations in ChatGPT, Perplexity and AI Mode
AI referral traffic is small but unusually good. Semrush's July 2025 study estimated the average AI search visitor is 4.4 times as valuable as the average organic visitor on conversion rate, on the logic that the model has done the comparing before the click (Semrush). Treat the multiplier as a direction rather than a constant, since it varies a lot by site. It still means a citation is worth chasing when the click count looks trivial.
What gets cited is not the same across engines. Semrush's three-month study of 230,000 prompts found Reddit, Wikipedia, LinkedIn, Forbes and Medium were the most cited domains overall. ChatGPT leaned on Wikipedia and Reddit, Perplexity on Reddit and LinkedIn, and Google AI Mode on LinkedIn, YouTube and Reddit (Semrush).
For a B2B site, the things that reliably move citations in our experience:
- A direct answer in the first two sentences under a heading that matches the sub-question, so the passage can be lifted cleanly.
- Original numbers, benchmarks or definitions that only you can be the source for. A model has to cite something when it quotes a figure.
- A consistent product description, in your own words, repeated across your site and your third-party profiles, so the model has one version of you to remember.
- Comparison pages that mention competitors by name, because the fan-out queries mention them.
We go deeper on passage structure in how to structure content so LLMs cite you. The short version is that AI answers quote passages, not pages, so the unit of optimisation has shrunk.
Reddit and LinkedIn are part of the B2B research phase now
Look at that Semrush citation list again. Two of the top three most cited domains are places where people talk about vendors rather than vendor sites. AI Mode cited LinkedIn in nearly 15% of responses in that study, and Reddit topped Perplexity.
That matches how B2B buyers behave. They search "[category] reddit" to get past the marketing, and they check whether anyone at the vendor says something worth reading on LinkedIn. Both surfaces then feed the AI answers the next buyer sees.
The playbook is unglamorous. On Reddit, find the three or four subreddits where your category is discussed, answer questions properly under a real account, and disclose who you are. Our Reddit SEO guide covers what gets accounts banned and what gets threads ranking, and we run Reddit marketing for clients who want it done with some discipline.
On LinkedIn, the asset is a named person publishing opinions with evidence. Founder and specialist posts get quoted; company-page reposts of blog links do not.
Measuring B2B SEO beyond clicks
If your only metric is organic sessions, the last eighteen months look like failure, and the picture is wrong. Here is the set we report on.
Impressions, split by surface. Google's Search Console AI features report, which rolled out globally at the end of August 2026, shows impressions from AI Overviews and AI Mode by page, country and device, but no clicks (Search Engine Land). It is the closest thing you have to a citation count from Google itself. Watch the trend, not the absolute number.
One caveat on baselines: Google's removal of the num=100 parameter in September 2025 wiped bot impressions out of Search Console. In one analysis of 319 properties, 87.7% lost impressions and 77.6% lost unique ranking keywords (Search Engine Land). Any year-on-year impression comparison that straddles that date is comparing two different definitions. We covered why impressions and clicks have drifted apart in the great decoupling.
Branded search. The 6sense data says most buyers arrive at a preferred vendor before they talk to anyone. Branded queries, and brand plus category queries ("[brand] pricing", "[brand] vs"), are the earliest visible sign that the invisible research is going your way. Track them weekly.
Self-reported attribution. A free-text "how did you hear about us" field on the demo form catches ChatGPT, a Reddit thread and a colleague's Slack message, none of which last-click sees. It is the cheapest attribution upgrade a B2B company can make.
Pipeline, not leads. Tie organic-sourced and organic-influenced deals to the CRM and report qualified pipeline and closed revenue by content group. A comparison page that touched four deals beats a guide that brought a thousand sessions from people who will never buy.
Technical basics B2B sites get wrong
None of this works on a site that hides its best material. The same mistakes come up in almost every B2B audit we run.
Gated content that should be open. The whitepaper behind a form is invisible to Google and to every AI model, and the buyer who wanted it will get a competitor's version in the AI answer instead. Our rule: gate the template, the calculator or the dataset, and publish the argument.
Thin solution pages. "Solutions for finance teams" with three paragraphs of adjectives and a demo button cannot be retrieved for any fan-out question. Give it the workflow, the screenshots and the proof, or fold it into a page that has them.
Pricing inside a JavaScript widget that renders nothing crawlable. Publish a plain-HTML version of the tiers alongside it.
Documentation on a noindexed subdomain. Your docs are the most specific, most citable text you own. Make them crawlable and link to them from product pages.
Duplicate persona pages. Six near-identical "for [industry]" pages with the noun swapped. Google collapses them and models learn nothing distinctive from any of them. Write three that are genuinely different.
A technical SEO audit surfaces these in a day. Fixing the content underneath takes the rest of the quarter.
A 90-day B2B SEO plan
This is the order we run it in for a new B2B client. It assumes one marketer and some access to a product person and a salesperson.
| Weeks | Focus | Output |
|---|---|---|
| 1 to 2 | Audit and baseline | Technical audit; list of gated assets to open; Search Console baseline for clicks, impressions (web and AI features), branded queries; "how did you hear about us" field live on forms |
| 3 to 4 | Buying-committee keyword map | Interview two salespeople and one customer; list the questions each committee role asks; map them to fan-out sub-questions per commercial topic; score by CPC and deal relevance rather than volume |
| 5 to 7 | Bottom-funnel build | Pricing page, three "vs" pages against the competitors sales hears most, one "alternatives to [incumbent]" page, integration pages for the top five integrations |
| 8 to 9 | Problem-led and product-led content | Four pieces that start from a real customer symptom and show the product solving it; one original benchmark or dataset only you can publish |
| 10 to 11 | Distribution and citations | Named-person LinkedIn cadence; Reddit presence in the category subreddits; consistent product description pushed to G2-style profiles and partner directories; internal links from docs to product pages |
| 12 to 13 | Measure and re-plan | Compare AI features impressions, branded query trend, self-reported attribution and organic-influenced pipeline against the baseline; decide the next quarter's topics from what got cited and what got forwarded |
Ninety days will not finish a ten-month sales cycle, so do not judge the plan on closed revenue at day 90. Judge it on whether the right pages exist, whether they are being cited, and whether branded and comparison queries are moving.
The short version
B2B SEO in 2026 is two disciplines wearing one name. On research queries, the goal is to be the source the AI answer cites, because the click is mostly gone. On commercial queries, the goal is still the click, and the pages that earn it are pricing, comparisons, alternatives, integrations and honest use-case pages.
Filter keywords by who is asking and what they need to decide, build the pages that answer the fan-out questions, publish the arguments you used to gate, and measure impressions, brand and pipeline rather than sessions.
If you want a second pair of eyes on which of those pages you are missing, that is what our content strategy work is built around, and a free SEO review is a sensible place to start.
