Summary: When AI engines answer software buyer queries, the sources they cite most are category-specific databases, vendor documentation and practitioner writing — not general tech press. In the Enterprise Software category of the Machine Relations Index, the most-cited source for "best tools" questions is a specialist ERP database cited in 34 of 101 monitored answer runs. TechCrunch was cited in 6 of 596 enterprise software news runs. G2 is cited in 456 of 16,039 monitored runs overall and is observed by all six engines.
Last updated: September 25, 2026
What we measured, and what we cannot measure #
The Machine Relations Index does not publish a "SaaS" vertical, and this page will not invent one. The nearest published category is Enterprise Software, and this analysis is built on that category alone.
The Index keeps each question shape separate, with its own run denominator, and does not pool rates across shapes. Neither does this page. A source that leads "best tools" questions may be absent from "how buyers choose" questions, and averaging those two together produces a number that describes nothing. Every figure below therefore carries the shape it was measured in and the denominator it was measured against.
Release mri_score_v2.0+2026-09-20+47973f373a20, window 2026-05-10 to 2026-09-20, 16,039 monitored answer runs across six engines (ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews, Perplexity). Google AI Mode and AI Overviews are monitored separately because they are separate surfaces with separate behaviour, and because AI Mode became a default search experience rather than a lab feature during this measurement period (Google, 2025). The published Enterprise Software category covers seven question shapes:
| Question shape | State | Observed runs | Run dates | Distinct cited domains |
|---|---|---|---|---|
| Best tools | published | 101 | 7 | 167 |
| How buyers choose | published | 99 | 7 | 172 |
| Is it worth it | published | 107 | 7 | 194 |
| News-driven citations | published | 596 | 53 | 1,375 |
| Problem-first research | published | 106 | 7 | 258 |
| Top lists | published | 114 | 7 | 264 |
| Comparisons | published | 114 | 7 | 186 |
Who AI engines actually cite for software buyer questions #
"Best [category] software" questions — 101 observed runs #
| Rank | Source | Source role | Cited runs | Rate |
|---|---|---|---|---|
| 1 of 167 | ERP Research | Market and company database | 34 | 33.66% |
| 2 of 167 | Top10ERP | Other observed source | 21 | 20.79% |
| 3 of 167 | Forbes | Editorial publication | 14 | 13.86% |
| 4 of 167 | Encoursa | Other observed source | 12 | 11.88% |
| 5 of 167 | Domo | Other observed source | 11 | 10.89% |
| 6 of 167 | ERP-Pilot | Other observed source | 11 | 10.89% |
| 7 of 167 | G2 | Market and company database | 11 | 10.89% |
| 8 of 167 | IFS | Other observed source | 11 | 10.89% |
| 9 of 167 | Zapier | Vendor-owned source | 11 | 10.89% |
| 10 of 167 | Doss | Other observed source | 10 | 9.90% |
PR Newswire is cited in 4 of these 101 runs (3.96%), rank 48 of 167. TechCrunch, VentureBeat, Business Insider and PCMag are not observed in this segment at all.
"How buyers choose" questions — 99 observed runs #
| Rank | Source | Source role | Cited runs | Rate |
|---|---|---|---|---|
| 1 of 172 | ERP Research | Market and company database | 28 | 28.28% |
| 2 of 172 | Rand Group | Other observed source | 18 | 18.18% |
| 3 of 172 | TechTarget | Editorial publication | 18 | 18.18% |
| 4 of 172 | Deloitte | Analyst and consulting research | 12 | 12.12% |
| 5 of 172 | Granuler | Other observed source | 12 | 12.12% |
| 6 of 172 | NetSuite | Vendor-owned source | 12 | 12.12% |
| 7 of 172 | Bista Solutions | Other observed source | 10 | 10.10% |
| 8 of 172 | Top10ERP | Other observed source | 10 | 10.10% |
| 9 of 172 | SAP | Other observed source | 9 | 9.09% |
| 10 of 172 | Digital Advisory | Other observed source | 8 | 8.08% |
News-driven questions — 596 observed runs across 53 dates #
| Rank | Source | Source role | Cited runs | Rate |
|---|---|---|---|---|
| 1 of 1,375 | Community and social platform | 66 | 11.07% | |
| 2 of 1,375 | Zylo | Vendor-owned source | 58 | 9.73% |
| 3 of 1,375 | Salesforce | Vendor-owned source | 50 | 8.39% |
| 4 of 1,375 | SaaStr | Editorial publication | 48 | 8.05% |
| 5 of 1,375 | Community and social platform | 47 | 7.89% | |
| 6 of 1,375 | BetterCloud | Vendor-owned source | 46 | 7.72% |
| 7 of 1,375 | Maxio | Vendor-owned source | 44 | 7.38% |
| 8 of 1,375 | Monday.com | Vendor-owned source | 43 | 7.21% |
| 9 of 1,375 | SaaS Mag | Editorial publication | 43 | 7.21% |
| 10 of 1,375 | Software Equity Group | Market and company database | 38 | 6.38% |
This is the segment where general tech press should look strongest, because it is the segment built from news events. It does not. TechCrunch is cited in 6 of 596 runs (1.01%), rank 190 of 1,375. VentureBeat is cited in 3 runs (0.50%), rank 400. PR Newswire is cited in 1 run (0.17%), rank 1,076 of 1,375. Forbes is the strongest general-press source here at 22 runs (3.69%), rank 24.
The other four shapes, in one line each #
- Comparisons (114 runs): ERP Focus leads at 23 runs (20.18%), then LinkedIn at 18 (15.79%) and TechTarget at 14 (12.28%).
- Problem-first research (106 runs): Reddit leads at 14 runs (13.21%), then Zapier at 12 (11.32%).
- Top lists (114 runs): the leaders are procurement specialists — ProcureDesk at 19 runs (16.67%), Procurify at 16 (14.04%).
- Is it worth it (107 runs): ERP Research leads at 26 runs (24.30%), then NetSuite at 18 (16.82%).
Across all seven shapes, no general technology publication leads a single one.
Review platforms are among the most-cited sources in the Index #
G2, Capterra and Product Hunt are often left out of AI visibility plans in favour of tech-press PR. The Index shows engines citing review platforms heavily.
| Surface | Cited runs (of 16,039) | Rate | Engines observing it | Confidence | Standing |
|---|---|---|---|---|---|
| G2 | 456 | 2.84% | 6 of 6 | Confidence A | rank 8 of 22,320 observed domains |
| Capterra | 86 | 0.54% | at least 5 of 6 | Confidence C | rank 89 of 22,320 |
| Product Hunt | 6 | 0.04% | at least 2 of 6 | Collecting | below the evidence floor |
| PR Newswire | 182 | 1.13% | 6 of 6 | Confidence B | — |
| TechCrunch | 164 | 1.02% | 5 of 6 (Gemini absent) | Confidence B | — |
| VentureBeat | 46 | 0.29% | 6 of 6 | Confidence C | — |
Engine-breadth note, updated September 26, 2026. Three of the six monitored engines had dated collection gaps inside this release's window, not the two this note reported until today. Google AI Overviews recorded no citations from August 15 to September 24, 2026; Google AI Mode none from September 14 to September 24; and Gemini none on 22 intermittent days — May 11 to 15, then scattered single days from August 11 to September 16. ChatGPT, Claude and Perplexity collected throughout; one Claude date was flagged by the same scan and carries exactly one answer run, which is a scan artifact rather than a collection gap. An engine count is a count of engines observed citing a domain, so a gap can only lower it, never raise it. Where any of the three gapped engines is among the engines absent from a domain, the count above is stated as a floor — "at least N" — because this release cannot tell an engine that passed the domain over from one that had no chance to see it. A count is exact only where every absent engine is ChatGPT, Claude or Perplexity, and then the absent engine is named. Cited runs, days cited, confidence states and standings are computed on the full run set and are unaffected. Citation rates are not: a blind run stays in the denominator while contributing nothing to the numerator, so every rate on this page taken over a window spanning these gaps is a floor and understates.
G2 is cited two and a half times as often as PR Newswire and TechCrunch, by every engine the Index monitors, at the highest confidence state the release publishes. Inside Enterprise Software it is cited in 11 of 101 "best tools" runs (rank 7 of 167), 30 of 596 news-driven runs (rank 14 of 1,375) and 7 of 114 "top lists" runs (rank 22 of 264). Capterra is cited less often than G2 but is comfortably present. Product Hunt is the only one of the three that sits below the release's evidence floor, and it is reported as collecting rather than as a rate.
The measurement point is worth carrying into any vendor report: a tracker that only watches a hand-picked list of publications returns no citations for a surface it never watched, and "no citations in the tracker" is not the same as "not cited by AI engines." Measuring every cited domain on one denominator is what separates the two.
What this means for SaaS brands #
The measured picture is less flattering to public relations and more useful to operators.
Category-specific sources dominate buyer-intent answers. The sources leading "best tools", "how buyers choose", "top lists" and "comparisons" are specialist databases, procurement and ERP publications, and vendor documentation — sources that exist to answer exactly the question being asked. A model assembling a shortlist reaches for the page that already contains a shortlist with reasons attached.
Vendor-owned pages are cited, heavily. Zylo, Salesforce, BetterCloud, Maxio, Monday.com, Zapier and NetSuite all appear in the tables above. Your own documentation, pricing pages and comparison pages are candidate sources, not just conversion surfaces.
Review platforms belong in the plan. G2 is among the most-cited domains in the entire Index. Being accurately represented there, with current pricing, current category placement and substantive review text, is AI visibility work, not only funnel work.
General tech press still matters, but not for this. Forbes is the one general publication that holds a top-three position in a buyer-intent shape (14 of 101 "best tools" runs). TechCrunch and VentureBeat are near-absent from software buyer questions in this window. Coverage in them buys reach, recruiting signal and credibility with humans; on the evidence here it does not buy citation volume in software buyer answers.
Communities are load-bearing. Reddit leads problem-first research questions (14 of 106 runs) and is cited in 2,016 of 16,039 monitored runs overall (12.57%), the most-cited domain in the Index. LinkedIn leads news-driven enterprise software questions at 66 of 596 runs.
The structural lesson from entity concentration in AI citations still holds — citation share concentrates fast — but for software buyers it concentrates on category specialists, review platforms and vendor documentation.
What the independent studies say, and where they disagree with us #
Third-party citation studies of AI search generally agree on the direction and disagree on the size, which is the normal state of a young measurement field. They are worth reading alongside this page rather than instead of it.
Quoleady's 10,000-citation study found that 50% of AI citations for high-intent SaaS prompts go to list-format articles — "Top X tools", "Best Y software", "X alternatives to [competitor]" (Quoleady, March 2026). Our tables agree on format and differ on host: the list-format pages being cited are mostly on specialist databases and vendor sites, not on general editorial publications.
BeVisibleIQ's 75-query study found that 80% of cited pages at the purchase-decision stage contain specific numbers — pricing, benchmarks, ROI figures — rather than star ratings (BeVisibleIQ, April 2026). That is consistent with what leads our buyer-intent shapes.
GrackerAI's cross-platform analysis found Reddit at 40.1% of cross-platform citations for B2B queries (GrackerAI, February 2026). We measure Reddit at 12.57% of monitored answer runs. Both can be true — they count different things over different query sets — and the shared finding is that a community platform outranks every publication either study names.
Yang et al. analyzed more than 366,000 citations embedded in AI responses across three providers and found citation concentrating on a relatively small set of sources rather than spreading (Yang et al., 2025). Muck Rack's Generative Pulse research found earned media consistently driving citations across AI systems (Muck Rack Generative Pulse, May 2026). Google's own guidance to publishers is that AI search still depends on crawlable, useful content (Google Search Central, 2025), and the original generative-engine optimization research found that adding statistics and credible citations improves visibility (Aggarwal et al., 2024). Meltwater's 2026 AI visibility reporting describes the same tightening of citation around trusted sources (Meltwater, 2026) — and for software buyer questions, the sources engines trust are the category specialists rather than the general press.
Machine Relations research on content structure and AI citation behavior covers what to do with a page once the placement exists.
Frequently Asked Questions #
Which sources do AI engines cite most for software buyer queries? #
In the Enterprise Software category of the Machine Relations Index, specialist databases and vendor sources lead every buyer-intent question shape. ERP Research is cited in 34 of 101 "best tools" runs and 28 of 99 "how buyers choose" runs. G2 is cited in 11 of those 101 "best tools" runs. LinkedIn leads news-driven questions at 66 of 596 runs. Forbes is the strongest general publication, cited in 14 of 101 "best tools" runs.
Do G2 and Capterra appear in AI citation pools? #
Yes. G2 was cited in 456 of 16,039 monitored answer runs (2.84%) in the 2026-05-10 to 2026-09-20 window, observed by all six engines at Confidence A, ranking 8 of 22,320 observed domains. Capterra was cited in 86 runs (0.54%) by at least five engines; Google AI Overviews is the one absent from its list, and that engine has a collection gap in this window, so five is a floor. Product Hunt, at 6 cited runs, sits below the evidence floor and is reported as collecting.
Is TechCrunch coverage worth pursuing for AI citations? #
Not on this evidence, if citation volume in software buyer answers is the goal. TechCrunch was cited in 6 of 596 enterprise software news-driven runs (1.01%), rank 190 of 1,375, and does not appear in any buyer-intent shape in this window. Across the whole Index it was cited in 164 of 16,039 runs, and it is the only source in that upper group not observed by all six engines: zero Gemini citations across the window. Gemini collected every day of that window, so its absence is a measurement rather than a collection gap, and TechCrunch's 5 of 6 is exact.
Does wire distribution drive SaaS AI citations? #
No. PR Newswire was cited in 4 of 101 enterprise software "best tools" runs and 1 of 596 news-driven runs, and in 182 of 16,039 runs across the whole Index (1.13%) — less often than G2 at 2.84%.
Is Reddit relevant for SaaS AI citations? #
Yes, more than any publication in this study. Reddit was cited in 2,016 of 16,039 monitored answer runs (12.57%), the most-cited domain in the Index, and it leads problem-first enterprise software research questions at 14 of 106 runs.
Why does this page not publish a single SaaS citation rate? #
Because the Index does not support one. There is no SaaS vertical in the release; the nearest published category is Enterprise Software, and each question shape inside it keeps its own denominator. Pooling those shapes would produce a single number that describes no question anyone asks. The earlier version of this page published pooled cumulative counts with no denominator at all, which is how it ranked sources by watch time instead of citation.
machinerelations.ai is the research and reference site for Machine Relations — the discipline of building AI-engine authority through earned media, entity clarity, and citation architecture. Related research: Top Enterprise AI Publications for AI Search 2026
Updated 2026-09-25: the engine-coverage column now distinguishes exact counts from floors, per the engine-breadth note above. Figures reflect the current Machine Relations Index methodology; see the MRI methodology and update log.