Why ranked-by-ad-spend B2B directories are dying
Click-through rates on G2 category leader pages dropped 31 percent in 2024. Capterra organic traffic share fell from 38 to 24. Buyers got AI. Vendors got fatigue. Here is what comes next.
Buyers don''t believe directory rankings anymore. The data shows it. Click-through rates on G2''s category leader pages dropped 31 percent in 2024 according to SimilarWeb. Capterra''s organic share of B2B software discovery search traffic fell from 38 percent to 24 percent over the same period. The tools that ride on top of those directories (sponsored placement, paid review acquisition, ranking auctions) are starting to look like Yelp''s last decade.
What changed?
Three things. Buyers got AI. Vendors got pay-to-rank fatigue. And the ranking math behind the directories was always indefensible.
Buyers got AI
A B2B buyer in 2025 doesn''t search "best CRM for agencies." She asks Claude or ChatGPT, "I run a 40-person digital agency. Our reps spend most of their time on outbound. We use Slack and Notion. What CRM should we be looking at?" The AI returns a reasoned, opinionated, three-option answer. Then the buyer reads two product sites, books one demo, and signs.
The directory step in the funnel was always information arbitrage. The directory had structured data the buyer didn''t. AI removed the arbitrage. The buyer can now ask for the structured answer directly.
Vendors got pay-to-rank fatigue
Run the math on a typical category. To stay in the top three of G2''s "CRM" category, a vendor pays roughly $80K to $200K a year in profile sponsorship. To do the same on Capterra, similar. To stay competitive on review volume, more in incentive programs that buy reviews from real customers (legal, but distasteful). All of this is pure overhead that comes out of margin. Vendors are starting to ask why.
The first wave of vendors to abandon paid placement was always going to be the ones with strong organic growth and product-led acquisition. They didn''t need the directory anyway. The second wave is happening now. Mid-market vendors with good products but mediocre paid-acquisition economics are pulling out, redirecting that budget to founder content, partner programs, and direct buyer relationships.
The ranking math was indefensible
Categories on G2 are scored by a formula that mixes review count, review recency, and "satisfaction" derived from those reviews. Reviews are mostly written by free users, incentivized through gift cards, or solicited at moments of high satisfaction (right after onboarding, before the bill comes). The result is a leaderboard that favors vendors with good incentive programs over vendors with good products. Buyers who have been through one or two procurement cycles figure this out. They stop trusting the leaderboard.
What replaces it
A few things, in parallel. AI advisors that are honest about their reasoning. Buyer-side communities that share real implementation experience. Marketplaces that rank by buyer outcomes (renewal rate, contract size, NPS at month 12) rather than review volume. Direct vendor relationships through partner programs that pay buyers for outcomes, not for clicks.
What this means for vendors
Pay attention to where your buyers actually are. If your buyers are running deals through a directory in 2025, the directory is part of the cycle. If they''re not (and increasingly they''re not), the directory is overhead.
Build content the AI can find. AI advisors quote sources. They quote the sources they can index, parse, and trust. The vendor with a clear pricing page, a clear feature matrix, and a documented data security posture is the vendor the AI recommends.
Build relationships outside the funnel. Partner programs, affiliate flows, embedded distribution. The vendors winning in 2026 are the ones who don''t need the directory.
What this means for buyers
Stop using the leaderboards as a primary filter. Use them as a sanity check after you''ve already narrowed the list through AI recommendations and one or two trusted humans. The directories are still useful for certain things (review volume as a signal of market presence, quick feature matrix scans). They''re no longer useful for the question they were built to answer.
The directory model is not dead. It''s just no longer the answer.