AI Overview Citations – What Appears for “Best [Product] Searches”: Original Data

When a buyer types “best CRM software” or “best fraud prevention software” into Google in 2026, the page they see is not the page B2B marketers spent the last decade optimizing for. The blue links are still there, but the answer has been pulled forward into Google’s AI Overview — a generated summary that names specific products, draws specific conclusions, and cites a specific set of sources right at the top of the result.

That citation set is the new B2B shelf. So we wanted to know who is actually on it.

We pulled the AI Overview citations for 100 SaaS shopping queries — the kind a buyer types when they are not loyal to a brand and are looking for options. Things like “best CRM for small business,” “best applicant tracking system,” “best fraud prevention software,” “best ocr software for developers,” and “best ehr software.” A balanced spread across sales, marketing, HR, finance, security, dev tools, AI/ML, and vertical SaaS.

We pulled the data via DataForSEO’s SERP API on May 6, 2026, US English, desktop, and we collected the top three references each AI Overview emitted in its source panel.

FULL DATA HERE

Here is what we learned.

99 of 100 SaaS shopping queries triggered an AI Overview

Before getting to the citations, the most important number is this one: AIO showed up for 99 of the 100 queries we tested. The single holdout — “best customer service software” — failed to produce an AI Overview across three separate pulls.

For commercial-intent SaaS queries, AIO is no longer an experiment, an A/B treatment, or a thing that “sometimes appears.” It is the default first impression. Anyone who is still treating it as a side surface is one or two product cycles behind the actual buyer experience.

We also confirmed that AIO presence is non-deterministic. Several keywords that returned no AIO on the first pull returned a full one on the second or third, and vice versa. Tracking AIO presence is a probabilistic exercise; one snapshot is not the truth.

Where the citations actually come from

Across the 99 keywords with an AI Overview, we collected 294 top-3 citations. Their distribution looks like this:

Source category	Citations	Share
Vendor – Listicle / Blog	150	51%
Forum / Community (Reddit, Quora)	43	15%
Vendor – Product / Category Page	33	11%
Video (YouTube)	28	10%
Analyst (Gartner / Forrester / IDC)	16	5%
Review Aggregator (G2 / Capterra / TrustRadius / Software Advice)	9	3%
News / Media	9	3%
Professional / Org / Gov	3	1%
Social / Pro Network	3	1%

A few of those rows deserve unpacking.

Finding 1 — Vendor listicles are the dominant format. By a lot.

More than half of all top-3 citations are vendor-published “best of” listicles. This is the now-familiar pattern: a vendor publishes a ranked guide on their own blog (“The 15 best applicant tracking systems in 2026”), conveniently includes themselves somewhere in the ranking, and Google’s AI Overview cites the listicle as a source.

What’s striking is how often the citing vendor is in an adjacent category, not the one being queried. A few real examples from our pull:

• Mercury (a startup bank) cited #1 for “best accounting software”
• Hightouch (a reverse-ETL tool) cited for “best customer data platform”
• Lever (an ATS vendor) cited for “best applicant tracking system” — listing itself on its own ranking
• CXL (a CRO training company) cited for “best A/B testing tools”
• Vellum (an LLM-ops tool) cited for “best AI voice agent platform”
• Orca Security cited for “best CI/CD tools”
• Saleshandy cited for “best cold email software” — a self-listicle

This isn’t accidental. Listicle SEO produced by adjacent vendors is currently the most reliable path to AIO citation for SaaS shopping queries. It works because vendors writing “best of” content tend to be thorough, well-linked, and crawlable, and because Google’s model treats the format as authoritative for “best [category]” intent.

The implication for B2B marketing teams is direct: a single well-built listicle on a category-adjacent topic is currently more useful as an AIO acquisition asset than a year of paid placement on a review site.

Finding 2 — Reddit is the single most powerful domain in B2B AIO

Reddit is cited in the top 3 for 38 of 99 keywords (38%) and shows up somewhere in the AIO citations for 45 of 99 (45%). When we look at which domain holds the #1 citation, Reddit holds 27 of those slots — more #1 placements than every other domain combined.

That is a structural shift in how B2B buying signals are being aggregated. Google is reading software shopping threads on Reddit and treating them as primary sources for product comparison. Anyone building a B2B GTM strategy that ignores Reddit visibility — community presence, founder participation, AMAs, monitored mentions — is not just missing a channel; they are missing the citation source most likely to anchor their product’s AI Overview narrative.

It also reframes what “social listening” needs to mean for SaaS marketers in 2026. Reddit is not a sentiment-analysis project anymore. It is a top-of-funnel indexed surface that gets directly translated into AI-generated buying recommendations.

Finding 3 — Review aggregators are mostly invisible

The combined footprint of every well-known review aggregator — G2, Capterra, TrustRadius, Software Advice, GetApp, PeerSpot, Business.com — across 100 SaaS shopping queries is 9 citations. Ninety of the 99 keywords have no review-aggregator citation in their top 3.

This was the most counter-intuitive finding for us. The B2B research stack has spent a decade convincing buyers — and vendors — that those sites are where comparison happens. Google’s AI Overview, at least for the queries we tested, doesn’t appear to agree. It’s pulling from vendor blogs, Reddit threads, and YouTube comparison videos far more aggressively than from structured review repositories.

A reasonable hypothesis is that aggregator pages are too database-shaped — a wall of star ratings and feature toggles — for the model to extract a clean recommendation from, while a well-written vendor listicle reads as a coherent answer. Whatever the cause, the implication is the same: dollars currently flowing into review-site profiles are not buying AIO real estate.

Finding 4 — Among analysts, only Gartner breaks through

Of the 16 analyst citations across the dataset, all 16 are Gartner. Forrester and IDC do not appear in the top 3 of a single keyword in our pull. Gartner Magic Quadrant pages and Gartner Peer Insights pages both appear; the format Google likes is the same listicle-shaped, comparison-oriented page that vendor blogs are also producing.

For a SaaS marketing team, this means analyst placements remain a useful signal — but only one of them is showing up on the new shelf.

Finding 5 — YouTube is the second most-cited content type after vendor listicles

YouTube videos appear in the top 3 for 24 of 99 keywords. The cited videos are almost always comparison or “I tested 5” walkthroughs from independent creators, not vendor channels.

This is the most underbuilt channel in the dataset. Vendor-owned YouTube content is rarely cited; creator-owned content is cited heavily. SaaS companies looking for an asymmetric move — high-output, low-saturation, cited-by-AIO content — should be talking to product-review YouTubers the way they talked to G2 reviewers in 2018.

What this means for B2B SaaS GTM

If the data above generalizes, four moves get more important and three get less so.

More important:

1. Owned listicle SEO — published on the vendor’s domain, on category-adjacent topics, well-cited and well-structured. This is the highest-ROI AIO acquisition asset right now.
2. Reddit presence — measured not as marketing but as indexed surface area. What threads show up for your category? Are you in them?
3. Creator partnerships on YouTube — for “I tested 5” formats specifically.
4. Gartner placement — still works; the others don’t seem to.

Less important than the budget suggests:

1. Premium G2 / Capterra / TrustRadius placements — for AIO citation purposes specifically. Still useful for buyer-stage validation, just not a path into the AIO panel itself.
2. Forrester and IDC analyst relations — at least for AI search visibility.
3. Mainstream tech press — only 3% of citations.

Methodology and caveats

We pulled SERPs for 100 SaaS shopping keywords via DataForSEO’s /v3/serp/google/organic/live/advanced endpoint with load_async_ai_overview=true, parameterized for US English desktop, on May 6, 2026. Each query that returned no AIO on the first pull was retried up to two more times to control for AIO non-determinism.

We then took the first three entries from ai_overview.references[] for each keyword — that is, the order Google places sources in the AI Overview’s right-rail panel, not deduped. Categorization is rule-based: known forum, video, news, analyst, and review-aggregator domains are labeled directly; remaining domains are classified as “vendor” and split between listicle/blog and product/category page based on URL and title heuristics.

A few honest limitations:

• One pull per keyword (with retries) is a snapshot, not a stable measurement. AIO content varies between pulls.
• US-only. Non-US AIO behavior, especially in EU markets where AIO is gated differently, will look different.
• Desktop-only. Mobile AIO appears to be more aggressive in some categories.
• Top-3 only. Some AIOs return 12–15 references; the long tail tells additional stories we did not analyze here.
• Categorization is heuristic. A small number of edge cases (e.g. sites that are simultaneously vendors and publishers) get assigned to the closest-fit bucket.

Total API spend for the dataset: ~$0.45.

The full dataset — every keyword, every top-3 citation, every URL, every domain category — is available alongside this article. If you replicate the pull and get materially different category shares, we’d like to hear about it.