Research into how AI systems cite sources for SaaS-related buyer questions is beginning to show a consistent pattern: source selection is concentrated, contextual, and dependent on the platform and prompt. Analyses associated with Kevin Indig and separate B2B SaaS research both suggest that the domains and content formats surfaced in AI answers can vary substantially by buyer stage. That matters for software companies treating AI visibility as a new distribution channel, but it also calls for care when interpreting individual datasets.
The narrow study description circulating around US ChatGPT citations in December 2025, four SaaS buyer-journey stages, and unique cited domains appears directionally consistent with this research theme. However, the supplied evidence does not tie that full methodology cleanly to one published dataset. The most accurate conclusion is that credible studies exist, while their scope, metrics, and engine coverage should not be merged into a single set of findings.
What the citation research indicates
Kevin Indig's analysis of how AI picks its sources examines patterns in AI citation behavior. The research, as summarized in the supplied material, points to a concentration of citations among a relatively small group of domains and to variation across AI engines and content formats. Search Engine Land's coverage of Indig's dataset similarly reports that roughly 30 domains account for a large share of citations, while the balance between first-party and third-party sources changes by platform and prompt type.
A separate BeVisibleIQ study examined 75 B2B SaaS buyer prompts across four engines. Its findings, according to the supplied research, also emphasize that buyer-stage context affects which sources and formats are cited, and that vendor websites and third-party sources do not appear at the same rate across engines.
Taken together, the studies do not support a simple rule such as "publish more vendor content to earn more AI citations." Instead, they suggest that visibility depends on the relationship among the query, the buyer's decision stage, the AI platform, and the available source material.
| Research stream | Scope described in supplied research | Relevant finding |
|---|---|---|
| Kevin Indig analysis | AI citation patterns, including variation by engine and content format | Citations appear concentrated among a relatively small set of domains. |
| BeVisibleIQ study | 75 B2B SaaS buyer prompts across four engines | Buyer-stage context and engine choice affect first-party and third-party citation patterns. |
| Narrow ChatGPT study description | US ChatGPT citations, December 2025, four buyer stages, unique domains | The description aligns with the broader theme, but is not cleanly attributable to one published dataset in the supplied material. |
Why methodology changes the interpretation
Citation studies can look comparable while measuring different things. A count of unique cited domains answers a diversity question: how many distinct websites were selected. Citation volume answers a different question: how often sources appeared. Neither metric alone establishes whether a domain is consistently influential, highly visible for one query type, or merely present in a broad source set.
The same issue applies to prompts and markets. Findings from US ChatGPT prompts cannot automatically be generalized to other countries, other AI engines, or all SaaS searches. Similarly, a dataset organized around buyer journeys can reveal useful differences between awareness, evaluation, and purchasing questions, but it cannot by itself represent every way a buyer researches software.
For readers assessing headline figures from this area, the practical checks are straightforward:
- Confirm whether the study measured unique domains, total citations, or both.
- Identify the AI engine, geography, prompt set, and collection period.
- Check whether prompts were grouped by buyer stage and how those stages were defined.
- Separate first-party vendor citations from independent editorial, review, community, and documentation sources.
- Avoid combining statistics from separate datasets as though they describe one unified benchmark.
Implications for SaaS teams and AI governance
For SaaS marketing and product teams, the emerging lesson is not that AI citation outcomes are fully controllable. It is that they should be measured with more precision. A vendor's own site may be relevant for product specifications, documentation, pricing, and implementation details. Third-party sources may be more relevant where a user seeks comparisons, independent validation, alternatives, or category guidance. The supplied research indicates that these distinctions can shift by engine and prompt type.
That has implications beyond content strategy. Enterprise teams increasingly rely on AI assistants during software research, which means citation behavior can influence the information reaching evaluators before they visit a vendor website. Governance programs should therefore distinguish between improving factual source material and attempting to engineer answers. Clear product documentation, maintained pricing information, consistent technical claims, and credible third-party coverage are more durable inputs than isolated optimization tactics.
Developer tooling can also help teams create an auditable process. Rather than tracking a single rank-like metric, organizations can log representative prompts, record cited domains, classify source types, and compare changes across engines over time. This approach does not prove causation, but it can reveal where official documentation is absent, where competitor or review sites dominate, and where a buyer-stage question produces inconsistent answers.
For enterprises adopting SaaS through AI-assisted research, the same caution applies in reverse. AI citations are useful starting points, not procurement evidence. Buyers should validate claims against vendor documentation, commercial terms, security materials, and their own technical requirements.
AI-assisted software discovery is changing where buyers encounter product information, but measurement choices determine whether a visibility program produces useful decisions or misleading dashboards. Scalevise helps teams build a repeatable view of how their brand and competitors appear in AI answers, connecting prompt-level evidence with practical content and governance priorities. Use the Scalevise AI Visibility GEO Checker to identify citation patterns worth investigating, focus resources on the questions that matter to buyers, and start an AI Visibility scan.
Frequently Asked Questions
What do ChatGPT citation studies measure?
They can measure different things, including unique domains cited, total citation occurrences, source types, or differences across prompts and engines. The metric must be checked before comparing results.
Does the supplied research show that ChatGPT cites the same SaaS sources for every question?
No. The supplied research indicates that citation behavior can vary by prompt type, buyer-stage context, content format, and AI platform.
Why does the difference between unique domains and citation volume matter?
Unique domains indicate the breadth of sources selected. Citation volume indicates how often sources appear. They answer different questions and should not be treated as interchangeable.
Can SaaS companies rely only on their own websites for AI visibility?
The research does not support that conclusion. It indicates that the balance of first-party and third-party citations varies by engine and prompt type.
Conclusion
The available evidence supports a meaningful research trend: AI citation patterns in SaaS-related prompts are concentrated and shaped by context. It does not support treating separate datasets as one universal ChatGPT benchmark. Teams that document methodology, track source types, and evaluate buyer-stage questions separately will be better positioned to interpret AI visibility without overstating what the data shows.
Top comments (0)