The SERP for “who does ChatGPT cite” is a wall of data and nothing else: Ahrefs’ monthly 50-domain table, Zerply’s reports, live trackers, listicles. Every one of them answers the same question — which domains — and none of them answers the only question a B2B SaaS founder actually cares about: so what does that list mean for my category? This piece leads with the data, then does the translation, including the part the data cannot support.
The data, sourced
The anchor dataset is Ahrefs’ monthly most-cited-domains study, September 2026 edition, updated September 2, 2026 (https://ahrefs.com/blog/most-cited-domains-in-chatgpt/). Method, in Ahrefs’ own terms: every domain ChatGPT cited across a broad sample of U.S. queries spanning all topics, ranked by mention share — a domain’s citations as a percentage of the summed citations of the top sources. Data comes from Ahrefs’ Brand Radar product, and the article is regenerated monthly from its data.
The top of the table (Ahrefs, September 2, 2026):
| # | Domain | Mention share |
|---|---|---|
| 1 | reddit.com | 16.8% |
| 2 | en.wikipedia.org | 7.0% |
| 3 | consumerreports.org | 3.7% |
| 4 | forbes.com | 3.1% |
| 5 | engineerfix.com | 2.8% |
| 6 | legalclarity.org | 2.8% |
| 7 | healthline.com | 2.5% |
| 8 | youtube.com | 2.5% |
| 9 | pmc.ncbi.nlm.nih.gov | 2.4% |
| 10 | walmart.com | 2.3% |
The rest of the top 50 runs through Wikipedia’s companions (dictionaries, reference sites), health and government primaries (PubMed, FDA, CDC, Mayo Clinic), consumer-reputation sites (Trustpilot 1.5%, BBB 1.2%, J.D. Power 1.1%), tech/product reviewers (Tom’s Guide, TechRadar, rtings), retail and product catalogs (Target, Home Depot, Lowe’s, Edmunds, KBB, Cars.com, the Apple App Store), and a few publishers (Wired, Vogue, Good Housekeeping). LinkedIn sits at 1.1%, ranked 38th.
Three things to note before the translation:
- Reddit alone holds about 17% of top-source citation share — more than double second-place Wikipedia at 7.0% (Ahrefs, Sept 2, 2026).
- The list is dominated by consumer and primary sources. No software review platform appears in the top 50. No G2, no Capterra, no Product Hunt, no developer docs.
- It is not a B2B dataset. It is all U.S. topics on one surface, one month. Which means the interesting conclusion is structural: if B2B citation worked like this list, no B2B software vendor could ever be cited anywhere in the top 50 — but we know that isn’t true, because category answers do name vendors. B2B citation happens in a narrower, per-category set of domains, and this list shows what those domains are types of.
That structural read is the whole argument that follows. The citation mechanics underneath — which queries even trigger citations at all — are worth knowing before you map anything: in Profound’s study of roughly 700,000 U.S. ChatGPT conversations, about 18% of conversations triggered at least one web search, and cited conversations averaged ~6 unique citations, with sources clustering by vertical (Profound, published February 3, 2026: https://www.tryprofound.com/blog/chatgpt-citation-sources/).
Five source types AI actually cites in B2B buying
Read the top 50 not as domains but as types, and map each type onto where B2B buying queries — category recommendations, “[X] vs [Y]”, pricing questions, “is [X] worth it” — actually get their citations:
1. Community and peer sources. Reddit (16.8%), YouTube (2.5%), Facebook (1.4%), LinkedIn (1.1%). The single largest type in the dataset by a wide margin. B2B mapping: the threads where your category is compared — r/[your-vertical], peer questions on LinkedIn, community Q&A. AI leans on this type the hardest of all, and it is the one type a vendor cannot publish into directly.
2. Review and reputation platforms. Consumer Reports (3.7%), Trustpilot (1.5%), BBB (1.2%), J.D. Power (1.1%). This is the largest type whose B2B analogs are entirely absent from this list — because G2, Capterra, Gartner Peer Insights and Clutch only matter inside specific categories, not across all topics. For B2B buying queries, the review-platform type is where “is X worth it” and vendor comparisons get their trust evidence.
3. Expert and product comparison media. Tom’s Guide (2.1%), TechRadar (2.1%), rtings (1.1%), NerdWallet (1.3%), Edmunds and KBB for cars. Independent writers comparing products within a category, often including pricing. B2B mapping: category roundup coverage, analyst and vertical-industry write-ups, “[X] vs [Y]” content — including your own comparison pages when they are written like answers rather than ads.
4. Primary and authoritative sources. Wikipedia (7.0%), the NIH/PubMed/FDA/CDC/health-cluster entries (collectively a large share), Forbes (3.1%). B2B mapping: category definitions, your own documentation, security and trust pages, and news coverage that establishes the facts an answer needs.
5. Official product and pricing catalogs. This is the underread row: Walmart (2.3%), Target, Home Depot, Lowe’s, Edmunds, Cars.com, the Apple App Store. ChatGPT cites official product and pricing pages when the question is “what does it cost, what does it include, where do I buy it.” B2B mapping: your pricing page, your product pages, your “contact sales” alternative. When a buyer asks a price question and your site offers no machine-readable answer, the citation goes to a review site or a competitor that does.
Where your brand sits in each: a five-question self-audit
Run these once, per source type. They take an afternoon:
- Community: search Reddit and LinkedIn for your category and your brand. In the threads where buyers compare options, is your product mentioned — and if so, favorably, and currently? (You cannot optimize this type directly; you can be a legitimate presence in it.)
- Reviews: do you have a complete, current profile on the 2–3 review platforms your category actually uses? How many reviews, how stale, and what do the recent ones say?
- Comparison media: who wrote the last three independent “[category] comparison” or “[competitor] alternatives” pieces that named your category? Are you in them?
- Primary/authoritative: do your docs, trust and pricing pages answer the factual questions an AI answer would need, in pages shaped like answers — with structured data and dates? The checklist version is in our B2B SaaS citation checklist.
- Official catalog: if a buyer asked “what does [your product] cost?” to ChatGPT, could an answer cite your own site — or would it have to cite someone else?
The audit’s shape matters: one brand rarely fails all five and rarely passes all five. The common failure pattern is the finding below.
The finding: B2B brands spend on the type AI doesn’t lean on
Here is the argument, from the data:
- The most-cited list shows citation concentration in peer, review, comparison and official-catalog types — with community the single largest source type.
- No B2B software review platform reaches the all-topics top 50, which is a scope artifact: those platforms matter per-category. But the artifact cuts the way you’d expect — B2B answers are more dependent on their category’s review, comparison and official sources than this consumer list shows, not less, because B2B answers have fewer alternative authorities to fall back on.
- Meanwhile, the standard B2B SaaS content budget goes into the one source type this data says AI leans on least: the vendor’s own thin blog — listicles, thought leadership, category explainers published to a domain with no third-party standing. That is content for Google’s long tail, and it is the type of page that never appears as a trust source in an AI answer about a category.
So the misallocation is legible in the data: brands spend on their own domain’s blog while being absent from the five types — community, reviews, comparisons, primary, official catalog — that AI answers actually draw on. The absence is not hypothetical; it is the difference between the types on the list and the types a typical B2B content plan funds. (What that absence looks like in referral terms is a separate question — the AI referral base overall is small even while growing fast, per the referral traffic numbers — which is exactly why presence in the right types matters: you are not buying traffic, you are buying the eligibility to be in the answer.)
The three cheapest presences to build
Ordered by cost-per-presence, assuming the audit showed the typical gaps:
- Finish your review-platform profile and invite real customers. A complete, current G2/Capterra/Clutch (or your category’s equivalent) profile with honest review volume is the single highest-leverage type-2 presence, and most of it is maintenance, not production.
- Make the official-catalog rows real: a machine-readable pricing page plus one honest comparison page per top competitor. This is where the Walmart/Edmunds rows apply to software: price and comparison questions need a citable official page, and the fixes are on your own site, so you control them.
- Be a legitimate presence in one community surface. Not a campaign — a maintainer who actually answers the category’s real questions where buyers ask them. Community is the largest type in the dataset and the least controllable; the only defensible version is genuine participation.
What this data can’t tell you
The honest limits, because the translation above depends on respecting them:
- It is category-blind. All-topics data shows types, not your category. The domains that cite for “CRM for professional services” are a different, narrower set than this table, and no published dataset gives them to you yet.
- It is one surface. ChatGPT only. Perplexity, Gemini and Claude have different citation mixes and different mechanics; the surface-by-surface comparison is the open question.
- It is a snapshot of one month, from one vendor’s pipeline. Ahrefs regenerates the table monthly from Brand Radar data; the mix moves month to month, and this is vendor data used with their labels, not an independent audit.
- It is not per-brand. “Who gets cited for your money queries” is not answerable from any public list. That question is per-query, per-surface, per-month — which is what the free quick check runs on your own query set, and what our methodology documents: the per-category version of this table, for your company.