Which sources do Google AI Overviews cite for B2B software queries?

What are the key findings?

Google AI Overviews show up on almost every B2B software buying query we tested, and they lean on three kinds of sources: vendor websites, YouTube videos and Reddit threads. The sources most marketers budget for, review platforms and analyst research, are a small slice.

How often does Google show an AI Overview for B2B software queries?

Nearly always. In our sample of 132 US B2B software queries with a current Ahrefs SERP snapshot, collected between 28 August and 10 October 2026, 122 showed an AI Overview, a rate of 92.4%.

We wrote 150 queries, 15 in each of 10 categories. 17 of them had no US snapshot in Ahrefs, which usually means very low search volume, and one had only a snapshot from February 2024, so we dropped both groups from the rate. That leaves 132 measured queries.

Two categories hit 100%: CRM (15 of 15) and accounting (15 of 15), along with sales engagement at 11 of 11. The lowest were cybersecurity (10 of 12, 83.3%), customer support (11 of 13, 84.6%) and HR and payroll (12 of 14, 85.7%). Those are small differences on small counts. No category came close to being free of AI Overviews.

Query pattern mattered more than category. Every "alternatives to a vendor" query with a snapshot showed an AI Overview (31 of 31), and so did 31 of 32 "X vs Y" comparisons. Queries that named an industry or company size, such as "payroll software for restaurants" or "bi tools for healthcare", were the least likely at 13 of 18 (72.2%).

The pattern makes sense if you think about what Google is answering. A comparison or alternatives query has a clear answer shape: a short list or a side by side. A vertical query often surfaces a mix of local providers, vendor landing pages and forum threads, and Google seems more willing to hand that one to the ten blue links.

How many sources does an AI Overview cite?

About six. Across the 122 AI Overviews, Ahrefs recorded 742 distinct cited URLs, an average of 6.1 and a median of 6 per AI Overview. The range ran from 1 to 17.

One cluster stands out: 42 AI Overviews (34%) list exactly three sources. That looks like the collapsed view, where Google shows three source cards and hides the rest behind a "show all" control. So treat the citation counts as a floor. Some AI Overviews almost certainly cite more pages than the snapshot captured.

A practical reading: if your page is the fourth best source for a query, it may be cited and still sit behind a click. Being one of the first three listed matters more than being listed at all.

Which types of sources do AI Overviews cite?

Software vendors first, then YouTube, then publishers. Of the 742 citations, 327 (44.1%) went to software company websites, 152 (20.5%) to YouTube, 99 (13.3%) to publishers and media, 77 (10.4%) to Reddit, Quora, forums and LinkedIn, 48 (6.5%) to agencies and consultants and 37 (5.0%) to review and listing sites.

Two categories we expected to see are empty. Analyst research got zero citations: every Gartner link in the data (16 of them) points at a Gartner Peer Insights review page, not a research note. Wikipedia also got zero. For buyer queries like these, Google does not reach for encyclopedic or analyst sources at all.

The vendor share needs one caveat. Zapier is a software company, so we counted its blog as a vendor site, and it is the most cited non video, non forum domain in the study with 22 citations in 22 AI Overviews (18%). If you would rather treat Zapier's app roundups as a publisher, the vendor share drops to 41.1%. Vendors still lead.

Who the vendor is matters too. On "best X" and "alternatives" queries the cited vendor is rarely the market leader. It is usually a smaller company that wrote a roundup: BIGContacts on "best crm software", TaskRhino and Plaky on project management queries, Salesgenie and Consulti on "apollo.io alternatives". Vendor content marketing is the main feedstock for AI Overviews in B2B software.

User generated content is the second story. YouTube appeared in 68 of 122 AI Overviews (55.7%) and Reddit in 57 (46.7%). Reddit also tends to be listed first: in 45 AI Overviews (36.9%) the first source in Ahrefs' record was a Reddit thread, more than any other single domain. The YouTube count is boosted by video carousels, where Google attaches several videos to one AI Overview. On "X vs Y" queries, 25 of 31 AI Overviews cited at least one video.

Which domains do AI Overviews cite most for B2B software?

YouTube and Reddit, by a wide margin. YouTube collected 152 citations and Reddit 65. Then the curve flattens fast: Zapier 22, Forbes 16, Gartner Peer Insights 16, G2 14 and Salesforce 10. No other domain reached 10.

The list is short at the top and very long at the bottom. 310 distinct domains earned at least one citation, and 231 of them (74.5%) earned exactly one. That long tail is good news for smaller companies. Getting cited does not require being a household name. It requires having the page Google wanted for that specific query.

Review platforms are present but small. G2 appeared in 14 AI Overviews, mostly as category pages ("Best Marketing Automation Software") and competitor alternatives pages. Capterra appeared once. TrustRadius did not appear at all.

Forbes Advisor deserves a note. All 16 of its citations are Forbes Advisor pages built on two templates, "10 Best X Software" and "X vs Y (2026 Comparison)", and the brand shows up across CRM, project management, HR and accounting. PCMag (9 citations) plays the same role with its tested roundups. Publishers that still run structured, regularly updated software roundups get cited. General news sites without that format barely appear in our data.

How does the source mix change by software category?

The mix shifts more by category than the overall numbers suggest. Vendor sites take 62% of citations in customer support and 64% in sales engagement, against 37% in accounting. Accounting is the most video heavy category: YouTube took 30 of its 108 citations (27.8%).

Publishers matter in categories where consumer finance media already covers the space. Accounting (20 of 108 citations, 18.5%) and HR and payroll (14 of 78, 17.9%) pull heavily from Forbes Advisor, CNBC Select, NerdWallet, Business News Daily and the US Chamber of Commerce. In customer support and sales engagement, publishers took only 2 citations each. Those categories have almost no independent media coverage left, so vendors and their blogs fill the gap.

Review sites peak in cybersecurity at 8 of 73 citations (11.0%), and 5 of those 8 are Gartner Peer Insights pages. Security buyers lean on peer reviews, and Google appears to follow.

If you sell in one of these categories, read your row before you read the averages. A sales engagement company that pours budget into earned media coverage is fighting for a slot that barely exists in its AI Overviews. An accounting software company without a YouTube presence is missing over a quarter of the citations in its space.

Do AI Overviews cite the vendor's own site on "X vs Y" queries?

Sometimes, and less than you'd think. Across 31 "X vs Y" AI Overviews, the two named vendors received 34 of 233 citations (14.6%). 23 of the 31 AI Overviews (74%) cited at least one of the two vendors, but only 9 cited both, and 8 cited neither.

Look at "salesforce vs hubspot". Both vendors publish a comparison page, and both pages ranked organically in positions 5 and 6 of the same snapshot. The AI Overview cited neither. It linked to Avoma, ZoomInfo's blog, Resonate, CRMSwitch, Forbes Advisor, a Reddit thread and three YouTube videos. "asana vs monday" went the same way: asana.com's comparison page ranked fourth organically and was not cited.

"Alternatives" queries are harsher still. In 31 AI Overviews for "alternatives to [vendor]", the vendor being replaced got 4 of 167 citations (2.4%), in 3 AI Overviews. Rival vendors got 92 citations (55.1%). The "X alternatives" SERP is owned by competitors: Xero and Wave on QuickBooks, SigNoz and OpenObserve on Datadog, Crisp and Decagon on Intercom.

Do pages cited in AI Overviews also rank in the organic top 10?

Less than a third of the time. For a sample of 40 queries with an AI Overview, 4 per category, we also pulled the organic results in the top 10 positions of the same snapshot. Of the 300 URLs those AI Overviews cited, 86 (28.7%) also appeared as an organic result. At the domain level the overlap rises to 117 (39.0%).

YouTube explains part of the gap. Only 3 of 78 cited YouTube URLs (3.8%) also ranked organically, because videos sit in their own SERP block. Remove YouTube and 83 of 222 cited URLs (37.4%) were in the organic top 10. That is still well under half.

The reverse holds too. Of the 297 organic URLs in those 40 top 10s, 86 (29.0%) were cited by the AI Overview above them. And in 10 of the 40 queries, the AI Overview cited no URL from the organic top 10 at all.

Google says AI Overviews may use a "query fan-out" technique, issuing related searches across subtopics to find supporting pages. That fits what we see. The AI Overview draws on a wider pool than the ten results under it, so a page on page two of the main query can still be the cited source if it ranks for one of the sub-queries.

Do AI Overviews only cite high authority domains?

No, though big domains take most of the volume. The median Ahrefs Domain Rating of the 310 cited domains is 72. Weighted by citations the median is 90, pulled up by YouTube (DR 99) and Reddit (DR 95). Excluding those two, the citation weighted median is 80.

The low end is real. 21.9% of cited domains have DR below 50, and 10.8% of all citations came from them. More useful for planning: 52 of 122 AI Overviews (42.6%) cited at least one domain with DR under 50. TaskRhino, a project management consultancy at DR 34, earned 4 citations. Five domains with DR under 10 were cited, on alternatives, comparison, small business and "top tools" queries.

Low authority sites win when they publish the specific page nobody bigger has written: a Gusto alternatives roundup on Finstackk (DR 2), a NetSuite alternatives page for manufacturers on Digit Software (DR 26), a Gong alternatives page from LeadHaste, a sales agency at DR 38. Specificity beats authority on long tail buyer queries.

Does the same domain get cited across categories?

Only a handful do. Of 310 cited domains, just 14 were cited in three or more of our 10 categories, and 7 in five or more: YouTube and Reddit (all 10), Zapier (8), G2 and Gartner Peer Insights (7), PCMag and Salesforce (5).

Everyone else is a specialist. Xero's 7 citations all sit in accounting. Rippling's 5 sit in HR and accounting. Ravetree's three citations are all on agency project management. For most B2B companies the competition for AI Overview slots is local to their category, which makes it measurable: a few dozen domains share most of the citations in any one space.

This is why share of voice by category is a more useful metric than a global authority score. Count which domains get cited across the 15 to 30 queries that matter to you, and you have a real competitor list for AI search.

What should B2B marketers do with this?

Start by measuring, then fix the gaps in the order the data suggests. These rules are written as tests you can run on your own query set.

To keep your tracking consistent, use a sheet with one row per query and these columns:

query, category, intent, check_date, aio_present (y/n),
cited_urls (in order), your_domain_cited (y/n), your_cited_url,
first_cited_domain, competitor_domains_cited, youtube_cited (y/n),
reddit_cited (y/n), your_organic_position, notes

Re-check monthly. AI Overviews change faster than organic rankings, and a single snapshot tells you less than a trend. Our guide to ranking in Google AI Overviews covers the on-page side: answer-first structure, specific claims and pages that can be quoted as a citation. For the wider picture across ChatGPT, Perplexity and Gemini, see AI search optimization.

How did we collect and classify the data?

We wrote 150 buyer intent queries, 15 for each of 10 B2B software categories: CRM, project management, HR and payroll, accounting, cybersecurity, marketing automation, BI and analytics, customer support, developer tools and DevOps, and sales engagement. Each category mixes six patterns: "best X software", "best X for small business", X for an industry or company size, "X vs Y" with two well known vendors, "top X tools" and "alternatives to [vendor]".

For each query we called the Ahrefs API v3 serp-overview endpoint with country US, filtered to AI Overview rows (ai_overview_found, ai_overview and ai_overview_sitelink). An ai_overview row means an AI Overview was present. Each ai_overview_sitelink row is one cited source. Ahrefs' snapshot dates for the AI Overview queries ran from 28 August to 10 October 2026, with a median of 30 September. We collected everything through the API on 12 October 2026.

When the filtered call returned nothing, we made a second call for the organic results to tell "no AI Overview" apart from "no snapshot". 10 queries had a current snapshot and no AI Overview, 17 had no US snapshot in Ahrefs and 1 ("bi tools for startups") had only a snapshot from February 2024. We excluded the last two groups from all rates.

Cleaning steps, in order:

• We removed duplicate URLs inside the same AI Overview. YouTube links that differ only by a timestamp or language parameter count once.
• Three citations were Google redirect links (google.com/goto). We assigned them to the domain named in the source title: Toptal, Pulumi and Clari.
• One AI Overview ("best help desk software for small business") contained a product list of 12 links whose titles name a vendor while the URL points to an unrelated page, such as a job listing. We kept them in the CSV with a flag and excluded them from every number in this report.
• Domains are registrable domains, so blog.hubspot.com and www.hubspot.com both count as hubspot.com.

We classified each of the 310 domains by hand into one source type: software vendor (a company that sells software, including its blog), YouTube, publisher or media (editorial sites, including Forbes Advisor, PCMag and niche software review publications), Reddit and other forums and social sites (Reddit, Quora, Spiceworks Community, GitHub discussions, LinkedIn), agency or consultant (service firms and implementation partners), review or listing site (G2, Capterra, GetApp, Gartner Peer Insights and software directories), analyst research, Wikipedia, or other.

Domain Rating comes from the Ahrefs public Domain Rating endpoint, read on 12 October 2026, for every cited domain. For the overlap analysis we picked 4 AI Overview queries per category, 40 in total, spread across patterns, and pulled the organic results within the top 10 positions of the same snapshot. Positions include SERP features, so each list holds 6 to 9 organic URLs. We compared URLs after removing "www", trailing slashes and tracking parameters.

All counts, rates and charts on this page come from that dataset. You can download it below and rerun any of them.

What are the limitations of this study?

• One snapshot per query. AI Overviews change between searches and between days. Each query here is one Ahrefs snapshot, and snapshot dates vary by up to six weeks across queries.
• US only, one location setting. Results in other countries, languages or logged-in states will differ.
• Citation counts are a floor. 42 AI Overviews show exactly three sources, which suggests the collapsed view. Hidden sources are not in the data.
• Order is Ahrefs' order. "First listed" means first in Ahrefs' record of the AI Overview, which may not match on-screen order on every device.
• Query set is ours. We wrote the queries to cover common buyer patterns. We did not weight them by search volume, so a low volume query counts the same as a high volume one.
• Classification is a judgment call. The Zapier decision alone moves the vendor share by three points. The CSV includes our type for every domain so you can reclassify.
• DR is current, not historical. Domain Rating was read after collection, not on each snapshot date.
• Overlap uses a sample. The organic comparison covers 40 of the 122 AI Overview queries.
• Wave 1. This is the first of several waves. Later waves will repeat the same queries so we can report change over time.

For other numbers on AI search adoption and click behaviour, see our AI search statistics roundup. To compare the tools that can run this kind of check for you, see the best AI visibility tools.

FAQ

How often do Google AI Overviews appear for B2B software searches?

Very often. In our sample of 132 US B2B software queries with a current Ahrefs snapshot, collected between 28 August and 10 October 2026, 122 showed an AI Overview, or 92.4%. Every "alternatives to a vendor" query had one, as did 31 of 32 comparison queries. Industry specific queries were the least likely, at 13 of 18.

Which websites do AI Overviews cite most for B2B software queries?

YouTube and Reddit lead by a wide margin, with 152 and 65 citations out of 742. Zapier, Forbes Advisor, Gartner Peer Insights, G2 and Salesforce follow. By type, software vendors' own sites earn the most citations (44.1%), then YouTube (20.5%) and publishers (13.3%). Review sites take 5.0% and analyst research none.

Do AI Overviews cite the vendor's own site for "X vs Y" queries?

Not often. Across 31 comparison AI Overviews, the two named vendors received 34 of 233 citations, or 14.6%. Only 9 AI Overviews cited both vendors and 8 cited neither, even when both vendors' comparison pages ranked on page one. Third party comparisons, YouTube reviews and Reddit threads took most of the slots.

Do you need to rank in the organic top 10 to be cited in an AI Overview?

No. In a sample of 40 B2B software queries, only 86 of 300 cited URLs (28.7%) also ranked in the organic top 10 of the same results page, and 10 AI Overviews cited no top 10 page at all. Excluding YouTube videos the overlap is 37.4%. Ranking helps, but Google pulls sources from a wider pool.

Do AI Overviews only cite high authority websites?

Big domains take most citations, but low authority sites still get in. The median Ahrefs Domain Rating of cited domains was 72, and 42.6% of AI Overviews cited at least one domain with DR below 50. Small sites win when they publish a specific page, such as alternatives for one niche, that larger sites have not written.

Can I download the data behind this study?

Yes. The CSV has one row per AI Overview citation, with query, category, intent, citation order, URL, domain, title, Domain Rating, source type and snapshot date. It also lists queries with no AI Overview or no snapshot. It is licensed CC BY 4.0, so you can reuse it with a link to this page.

Sources

Dataset: ai-overview-citations-b2b.csv (742 citation rows plus 12 flagged rows and status rows for the 28 queries without an AI Overview). Licensed CC BY 4.0: reuse it freely with a link to this page. Also archived on Zenodo (DOI 10.5281/zenodo.23311409), Hugging Face and GitHub.