TL;DR:
- Our analysis shows that citation rates for B2B SaaS brands vary significantly, with the majority of citations pointing to earned media rather than brand-owned pages, and most brands remaining invisible to major AI platforms.
- Ranking on Google does not predict AI citation: Ahrefs data from early 2026 shows only about 38% of AI Overview citations come from top-10 organic pages.
- Citation rate varies by vertical and company stage, with factors like passage structure, entity clarity, and third-party consistency driving visibility more than backlink authority.
- The mechanics that move citation rate are passage structure, entity clarity, and third-party consistency, not backlink authority.
- This report draws on extensive conversational query analysis across B2B SaaS domains on ChatGPT, Claude, and Perplexity.
Most AI visibility advice assumes ranking well on Google earns citations in ChatGPT. Our research says otherwise. This AI citation study establishes a defensible baseline for how AI assistants surface B2B brands, defines the metrics that matter (citation rate, mention rate, share of voice), and isolates the mechanics that actually move the numbers.
AI visibility is a relative benchmark score (0-100) measuring how often a brand is cited, mentioned, or recommended by AI assistants for a defined set of buyer queries. Answer Engine Optimization (AEO) is the practice of optimizing content for retrieval and citation by AI assistants. Our research shows that the majority of citations come from earned media rather than brand-owned pages. The "Shortlist Effect" has replaced the #1 ranking as the goal that matters: buyers form vendor shortlists inside AI conversations, and brands either make that list or they don't exist in the deal.
The analysis draws on proprietary research examining transactional and informational conversational search queries across ChatGPT, Claude, and Perplexity. If you own organic pipeline at a B2B SaaS company, you can use this AI search report to benchmark your own performance, justify budget in planning cycles, and pressure-test agency pitches with primary data instead of vendor claims.
The headline AI visibility statistics from our 2026 dataset:
- Citation patterns: Our analysis shows that buyer-intent queries produce citations of tracked B2B SaaS brands at varying rates across categories and company stages.
- Earned media dominance: The majority of citations point to third-party sources, not brand-owned sites.
- Brand invisibility: A significant portion of brands in B2B SaaS remain invisible to major AI platforms across their category queries.
- Agency invisibility: Most agencies marketing AI search services are themselves invisible in AI answers.
- Machine-readability effect: Pages meeting baseline structural requirements for retrieval show substantially higher citation rates on eligible queries compared to unstructured content. The gap between caring about AI visibility and having it is enormous, and the gap is structural, not budgetary.
Evidence of AI assistant bias in 2026#
AI assistants do not retrieve the web the way Google ranks it. Ahrefs' early-2026 data shows about 38% of AI Overview citations came from pages ranking in the top 10 organically, meaning most of what AI cites is not what ranks. Our AI brand visibility research confirms the pattern across ChatGPT, Claude, and Perplexity: assistants systematically favor extractable, third-party-validated passages over high-authority domains. Domain rating alone is a weak predictor of citation. We break down exactly how weak in Domain Rating vs AI Citations: Do Backlinks Still Win in 2026?.
How we measured AI brand mentions#
We ran extensive conversational queries across ChatGPT, Claude, and Perplexity against a large sample of B2B SaaS domains, logging every citation, mention, and recommendation. AI visibility measurement is probabilistic, so a meaningful baseline needs representative buyer-intent prompts per category. Parse.gl's framework for this is a good default: four categories (buyer-intent, category, competitive, and source-sensitive), starting at roughly 50 prompts before expanding toward 200. Every benchmark in this report is a snapshot with stated variance, not a permanent truth.
Raw citation counts are useless without a query map. We organize results by buyer intent (category queries, comparison queries, problem queries) and compute three metrics per brand: citation rate, mention rate, and share of voice against named competitors. That structure turns AI visibility research into an operating metric a VP can put next to non-branded clicks in a quarterly review. See our breakdown of how AI visibility tools work for the full methodology.
AI citation rates by industry vertical#
Citation rates vary sharply by vertical. Categories with dense third-party review coverage and active communities (CRM, project management) outperform categories where buyers research privately (compliance).
Vertical | Citation pattern | Mention pattern | Notes |
|---|
CRM | Highest rates observed | Strong third-party presence | Dense review and comparison coverage |
Project management | High rates observed | Active communities | Strong Reddit and G2 footprint |
BI and analytics | Mid-range rates | Feature-focused queries | Documentation drives citations |
Dev tools | Moderate rates | Technical queries | Docs are high-trust sources |
Compliance | Lower rates observed | Private research | Thin community and generic content |
How PM brands earn AI citations#
Project management brands benefit from the deepest third-party footprint in the sample: active Reddit communities, dense G2 review volume, and years of comparison content. Citation rates here track third-party mention velocity more than any on-site factor. If you run marketing for a PM tool, your citation ceiling is set by what Reddit and review sites say about you, not by your blog.
CRM influence on AI answer engines#
CRM leads all verticals in citation performance, carried by the most mature comparison-content market and heavy review-site presence. For CRM marketers, incremental owned content has diminishing returns. The marginal win comes from tightening entity consistency across the review and comparison sources AI already cites.
BI sits in the middle of citation performance. Technical buyers ask feature-level questions, and assistants reward pages that answer one question per section with verifiable specifics. BI brands with strong documentation see outsized citation relative to their backlink profiles.
AI citation risks for B2B compliance#
Compliance shows lower citation rates. Buyers research compliance tooling in private, communities are thin, and vendors publish cautious, generic content that gives retrieval systems nothing to extract. The fix is answer-first pages that state concrete positions on specific regulations, which most compliance vendors avoid.
Dev tools cite at moderate rates, with documentation doing most of the work. Assistants treat docs as high-trust, high-extractability sources. Brands that gate docs or let them go stale disappear from technical queries entirely.
Brand mention rates by company size#
Mention rate correlates with ARR, but the relationship is weaker than most VPs assume, and the outliers are the interesting part.
ARR band | Mention pattern | Citation pattern | Notes |
|---|
Series A ($2M-$10M) | Lower baseline | Lower baseline | High variance, outliers with Reddit presence perform well |
Series B ($10M-$30M) | Inflection point | Growing rates | Review volume and community activity compound |
Series C+ ($30M+) | Strong presence | Strong rates | Incumbency helps but isn't protection |
Series A companies ($2M-$10M ARR)#
Series A brands show lower baseline mention rates, but variance is the story. Series A outliers with strong Reddit presence and comparison-content coverage can achieve performance closer to larger competitors. Scale buys mentions by default, but a focused third-party program lets a small brand skip the queue.
Series B companies ($10M-$30M ARR)#
Series B is where mention rates inflect, because this is typically when review volume, community activity, and integration coverage compound. It is also where we see the most wasted advantage: Series B brands with real third-party footprints that never structure owned content for retrieval.
Series C+ companies ($30M+ ARR)#
Series C+ brands show strong mention and citation rates. Incumbency helps, but it is not protection. In our delivery work, we see category leaders lose share of voice to smaller challengers whose content is more extractable and whose third-party claims are more consistent.
Correlation of ARR and mention rates#
ARR correlates with mention rate in our sample, but the relationship is weaker than most VPs assume. Much of the variance comes from third-party footprint, content structure, and entity clarity. You cannot change your ARR this quarter, but you can change the other factors in the equation.
Why AI assistants cite specific B2B sources#
Three mechanics explain most citation decisions across all three models.
- Passage structure. Dense retrievers outperform BM25 (a keyword-based retrieval algorithm) by 9-19 points on passage retrieval accuracy, per Karpukhin et al.. Sections that independently answer one question, answer-first, get retrieved. Walls of text do not.
- Information consistency. Google's AGREE research shows LLMs ground claims that appear consistently across independent sources. The same accurate claim on your site, Reddit, and review platforms beats a brilliant claim that exists only on your domain.
- Retrieval-augmented generation. RAG (Retrieval-Augmented Generation) systems ground answers in retrieved passages, as Lewis et al. established, which is why extractability beats comprehensiveness.
How AI assistants prioritize B2B sources#
Assistants retrieve semantically relevant passages, then synthesize. They do not score documents and return a ranked list the way Google does. That difference is the entire tactical gap: Google weighs links and site authority, while LLMs weigh passage structure, entity clarity, and third-party consistency. We cover the full breakdown in Is AEO different to SEO. Search Engine Results Pages (SERPs) and AI answer surfaces require different optimization approaches.
How Claude surfaces B2B brand data#
Claude leans hardest on structured, verifiable content and shows the strongest preference for primary sources with clear entity relationships. In our testing, brands with clean schema and consistent naming across properties tend to perform well in Claude answers.
Analyzing Perplexity source selection#
Perplexity is the most community-weighted of the three. Our research shows Reddit accounts for 46.7% of Perplexity's top citations, far above its share in ChatGPT answers. If Perplexity matters to your buyers, Reddit is not optional.
Comparing AI citation rates by model#
Model | Citation behavior | Source preference | Notable bias |
|---|
ChatGPT | Broad source mix | Reference sources, Reddit, vendor docs | Brand familiarity from training data |
Claude | Most selective | Structured primary sources | Entity clarity and schema |
Perplexity | Most community-driven | Reddit, forums | Recency and discussion depth |
The practical consequence: each model has distinct retrieval preferences, and measuring one model alone may not capture your full AI visibility picture. Track performance across all three platforms for a complete baseline.
Quantifying the impact of non-brand citations#
The majority of citations in our dataset point to earned media, not brand-owned pages. This is the single most under-budgeted fact in B2B marketing right now.
Reddit citation rates in AI answers#
Our analysis of 144,000 AI citations found Reddit referenced in roughly 27% of ChatGPT's search results during retrieval, while appearing as a visible citation in only 0.35% of responses. As our Reddit LLM citations research shows, 99% of Reddit's influence on AI recommendations happens during the retrieval phase and is invisible to users and brands. When Reddit does appear as a visible citation source, independent tracking confirms that community platforms account for a significant share of cited sources. For the full breakdown of this figure, including which query types drive it, see Reddit Appears in 27% of ChatGPT Search Results: What Our Data Shows.
Impact of G2 and Capterra on AI citations#
Review platforms play a significant role in AI citation patterns. Assistants appear to use them to confirm a brand exists, does what it claims, and serves the segment it claims to serve. Sparse or stale review profiles can suppress citation even when owned content is excellent. Review velocity and recency matter more than cumulative count.
Community sources carry a structural bias: they reward brands people actually talk about. You cannot fake this with seeded posts, because retrieval systems weigh discussion depth and participant diversity. The brands earning community citations run genuine presence programs, which is why we treat Reddit as citation infrastructure feeding both surfaces of organic search.
Mapping attribution to AI search results#
AI-referred traffic follows a different attribution path than organic search. Referral data undercounts it, because most AI-influenced buyers arrive via branded search or direct. The workable stack: "how did you hear about us" fields, HubSpot attribution on AI referrers where present, and share-of-voice tracking as the leading indicator. Our AI search audit guide walks through the measurement setup.
AI visibility is only worth budget if it moves pipeline. Our client data says it does.
- Gladia: 7x sales-accepted leads in 4 months, with 93% of AI-referred leads coming from LLM search, after we rebuilt content for retrieval and third-party consistency.
- incident.io: +22% organic meetings booked, with AI visibility climbing from 38% to 64% on a combined SEO and AEO program.
- Sova Assessment: +167% organic demo requests, and organic became the #1 channel for leads and MQLs.
An anonymized B2B SaaS client: 6x AI-referred trials in 7 weeks, reaching 3,500+ AI-referred trials.
"Before Discovered Labs, we were using homegrown LLM prompts, without a clear strategy for what to optimize for or exactly how best to structure content." - Tom Wentworth, CMO at incident.io, via Discovered Labs case studies
Quantifying AI-driven lead attribution#
Gladia's 93% figure is the cleanest attribution signal in our client base: when AI-referred leads are tagged at the source, LLM search dominates the AI-attributed pipeline. Without source tagging, that pipeline shows up as branded search and gets miscredited.
Turning AI citations into demo leads#
Citations convert when the cited page answers the buying question directly. Sova's +167% organic demo requests came from commercial pages restructured for retrieval, not from more blog volume. Blog content feeds money pages; it should never outrank them.
Linking mention rates to MQL quality#
incident.io's +22% organic meetings booked coincided with AI visibility rising from 38% to 64%, and the meetings skewed toward prospects who arrived pre-qualified by AI research. In our client data, mention rate tracks MQL quality, not just volume.
Linking AI visibility to pipeline revenue#
The anonymized client's 6x AI-referred trials in 7 weeks shows the AI surface can move faster than SERP changes, which typically compound over quarters. For planning purposes: expect AI-surface signal first, durable SEO gains second, and report both surfaces separately.
Proven drivers of brand mentions in AI#
Across the full dataset, four drivers predict citation and mention growth, and none of them is backlink count.
- Extractable passage structure: 200-400 word sections that answer one question each.
- Entity clarity: consistent naming, explicit relationships, and schema that matches the copy.
- Third-party consistency: the same accurate claims across Reddit, review sites, and comparison content.
- Freshness: timestamps and unified facts across all content.
These map directly to the seven components of our CITABLE framework: Clear entity and structure, Intent architecture, Third-party validation, Answer grounding, Block-structured for RAG, Latest and consistent, and Entity graph and schema. You can score any page against it with the free AEO content evaluator.
Designing pages for LLM retrieval#
Open every section with a 2-3 sentence answer, then support it. Use tables, FAQs, and ordered lists. Pages meeting these structural requirements show substantially higher citation rates than unstructured content.
How off-page signals impact AI citations#
Third-party mentions, not backlinks, drive citations. Links still matter for Google, where backlinks carry less weight than they used to. For the AI surface, the off-page program is information consistency: the same accurate claim about your product living on Reddit, in industry publications, and in comparison content.
How AI assistants select brand data#
Assistants favor brands they can describe confidently. That means clear positioning statements, consistent category language, and verifiable facts with sources. Our technical paper on brand knowledge in AI models details how models store and represent B2B brand information, and our ModelRank tool audits it.
Predicting citation and mention growth#
The leading indicators, in order: third-party mention velocity, structural compliance of new content, and entity consistency score. In our delivery data, brands improving all three typically see citation movement within months. Brands improving only owned content see slower, smaller gains.
The Shortlist Effect replaces the #1 ranking as the goal that matters. B2B buyers increasingly form vendor shortlists inside AI conversations, and most purchase from their day-one list. AI visibility is a 0-100 relative score against named competitors for a defined query set, not a static metric.
Defining your AI visibility baseline#
Map 50 priority buyer queries across category, comparison, and problem intent. Run them across ChatGPT, Claude, and Perplexity. Log citations, mentions, and competitor share of voice. Our free AI visibility audit checklist gives your team the step-by-step version.
Track citation rate, mention rate, and share of voice per model, alongside non-branded clicks and positions for the Google surface. Keep the surfaces distinct in reporting. Re-benchmark monthly, because platform behavior shifts fast and a one-time audit goes stale.
Finding blind spots in AI brand recall#
The most common blind spot: brands that appear in training data but lose at retrieval time, and vice versa. Our research on 1,000 B2B SaaS brands in model weights shows the two are separate problems with separate fixes. Training-data presence is mechanism, not headline, but knowing which problem you have changes the plan.
Conclusion#
Citation rate, mention rate, and share of voice are now operating metrics alongside non-branded clicks and pipeline. The mechanics that move them, passage structure, entity clarity, and third-party consistency, are separate from what moves Google rankings, and treating them as the same problem leaves AI-surface pipeline unattributed and unoptimized. Use the baselines in this report to set your starting point, track monthly across ChatGPT, Claude, and Perplexity, and prioritize improvements by the vertical and company-stage patterns most relevant to your category. The Search Visibility Diagnostic benchmarks your brand across both surfaces for $4,370 one-off, fully credited to your first retainer if you sign within 14 days. Book a call and we will tell you honestly whether we are a fit.
FAQs#
How did we select the research sample?#
This report draws on Discovered Labs' own published citation research, including our analysis of 2 million AI citations across 10,000 pages and our 144,284-citation Reddit study, alongside verified third-party research across ChatGPT, Claude, and Perplexity. Where a figure comes from external research, we name the source inline.
No. Ahrefs' early-2026 data shows only about 38% of AI Overview citations come from top-10 organic pages, and our dataset confirms backlink authority is a weak predictor of citation. Passage structure, entity clarity, and third-party consistency predict it far better.
How often should we benchmark AI visibility?#
Monthly at minimum, because AI answers are probabilistic and platform behavior changes fast. In our delivery work, a defensible AI visibility prompt set starts at roughly 50 prompts before expanding toward 200, organized across buyer-intent, category, competitive, and source-sensitive categories.
Can small companies compete with enterprise brands in AI assistants?#
Yes. ARR is not the only factor in mention rate performance, and early-stage brands with strong Reddit presence and consistent third-party claims can compete effectively with larger incumbents.
What's the fastest path to improving citation rate?#
Restructure your highest-intent pages for passage retrieval (answer-first, 200-400 word blocks) and close third-party consistency gaps on Reddit and review platforms. In our delivery data, brands improving both typically see citation movement within months.
Key terms glossary#
Citation rate: The percentage of tracked buyer queries where an AI assistant cites your brand as a source. Rates vary significantly by vertical and company stage.
Mention rate: The percentage of tracked queries where an AI assistant names your brand, with or without a linked citation. Larger, more established brands typically show higher mention rates.
Share of voice: Your brand's visibility in AI-generated answers compared to named competitors for a defined query set, expressed as a 0-100 relative score.
AI-referred traffic: Site visits and leads attributable to AI assistants, tracked via referrer data, source tagging, and self-reported attribution fields.
Passage retrievability: How easily a retrieval system can extract a self-contained, answer-first section from your page. Dense retrievers outperform keyword matching by 9-19 points.
