HR AI Shortlists: 4 of the Top 10 Are Not Universal Sources
The first published HR software buyer-choice leaderboard looks stable at the category layer and fragmented at the engine layer. Four of its top ten sources are absent from at least one of the three engines the Index collected without a gap in this window, while the HR news list is universal across all three.
The first published HR software buyer-choice leaderboard does not mean every AI engine is reading the same shortlist.
In the Machine Relations Index release generated September 25, 2026 (mri_score_v2.0+2026-09-25+04fcb7fb8f29), the HR & Talent / how-buyers-choose segment sits above the evidence floor: 101 observed answer runs across seven run dates, with 235 cited source domains. That gives the category its first citable buyer-choice list.
The top ten is already published. The engine layer is less settled. Four of those ten domains are absent from at least one of the three engines the Index collected without a gap in this window — ChatGPT, Claude and Perplexity. ChatGPT misses four of the ten, Claude one, and Perplexity is present on every top-ten domain.
Three engines are missing from that count on purpose. The Index's own collection has dated holes on all three: no citations from Google AI Overviews on any monitored query for 41 of the window's 132 days, none from Google AI Mode for 11, and 22 intermittent blind days on Gemini. A domain being absent from one of those three lanes cannot be read as engine behaviour, so none of them carries an absence count on this page. Presence on them is real and is shown; absence is not measurable and is not published as one.
That does not say which engine cited which source inside this exact HR segment. The public release does not expose per-engine rates at segment level. It says something narrower and still useful: the first HR buyer-choice shortlist is made of sources whose whole-window engine reach is uneven.
The buyer-choice top ten, with engine reach
The HR & Talent / how-buyers-choose denominator is 101 observed answer runs. A domain's segment rate is the share of those runs in which the domain was cited at least once. Engine reach below is whole-window presence in the release: whether each engine cited that domain anywhere in the measured panel between May 10 and September 25, 2026.
The exact column counts the three engines collected without a gap through the window. The last two columns report whether each gapped lane was recorded citing the domain at all; a blank there is not an absence, because all three lanes have missing collection days.
| Rank | Source | Segment runs | Source role | Exactly measured engines present (of 3) | Absent from (exact) | Gemini seen (floor) | Google surfaces seen (floor) |
|---|---|---|---|---|---|---|---|
| 1 | Pin | 14 of 101 | Vendor-owned source | 3 of 3 | — | Yes | AI Mode |
| 2 | 13 of 101 | Community and social platform | 3 of 3 | — | Yes | AI Mode, AI Overviews | |
| 3 | 9 of 101 | Community and social platform | 1 of 3 | ChatGPT, Claude | Yes | AI Mode, AI Overviews | |
| 4 | Selenios | 9 of 101 | Other observed source | 2 of 3 | ChatGPT | — | — |
| 5 | Umantis | 9 of 101 | Other observed source | 2 of 3 | ChatGPT | — | — |
| 6 | US Tech Automations | 9 of 101 | Other observed source | 3 of 3 | — | Yes | AI Mode, AI Overviews |
| 7 | ClearCompany | 8 of 101 | Other observed source | 3 of 3 | — | Yes | AI Mode, AI Overviews |
| 8 | CVViZ | 8 of 101 | Other observed source | 2 of 3 | ChatGPT | — | — |
| 9 | Gem | 8 of 101 | Vendor-owned source | 3 of 3 | — | Yes | AI Mode |
| 10 | Lever | 8 of 101 | Other observed source | 3 of 3 | — | Yes | AI Mode, AI Overviews |
Six of the ten are cited by all three exactly measured engines. Four are not. Three sources that sit inside the first eight ranks — Selenios, Umantis and CVViZ — were recorded only on Claude and Perplexity anywhere in the window, on no gapped lane at all, and all three are still at collecting confidence, with 8 to 13 total citations each across the whole release. They are visible enough to enter the first HR buyer leaderboard, but there is not yet enough evidence behind them to treat them as durable engine infrastructure.
That is the engine-reach warning. A source can rank inside a category answer while still being unevenly distributed across engines — and thin enough that the distribution itself is provisional.
The news list is broader
The HR news-shaped list looks different. It has 608 observed runs across 53 run dates, so it is thicker, older and less fragile than the new buyer-choice segment.
Its top ten are Unleash, Landbase, Crunchbase, Rippling, SHRM, Outsail, HR Executive, People Managing People, HR Tech Feed and LinkedIn.
All ten are cited by all three exactly measured engines. One of them, HR Executive, was not recorded on Gemini — which, with 22 blind days on that lane, is not an absence we can publish. None is collecting-confidence overall. None is limited to two engines.
| HR & Talent top-ten list | Cited by all three exactly measured engines | Cited by two or fewer | Collecting-confidence domains in top ten |
|---|---|---|---|
| Buyer-choice / how buyers choose | 6 of 10 | 4 of 10 | 3 of 10 |
| News-driven citations | 10 of 10 | 0 of 10 | 0 of 10 |
The difference is not just editorial versus vendor-owned. It is maturity and distribution. HR news answers are drawing from sources with broad engine reach and heavy evidence behind them. HR buyer-choice answers just became citable, and the first list still contains narrow, thinly evidenced surfaces.
That is why the category read has to keep two layers separate. The buyer leaderboard tells you what sources the model cites for the buyer-choice question. Engine reach tells you whether those sources are visible across the panel or concentrated in a smaller set of engines.
The engine pattern
Across the buyer-choice top ten, missing-engine counts look like this. Only the three engines with gap-free collection through this window can carry an absence count.
| Engine | Top-ten HR buyer-choice sources absent from that engine's whole-window citation set |
|---|---|
| ChatGPT | 4 of 10 |
| Claude | 1 of 10 |
| Perplexity | 0 of 10 |
| Gemini | Not measurable in this window (22 blind days) |
| Google AI Mode | Not measurable in this window (11 blind days) |
| Google AI Overviews | Not measurable in this window (41 blind days) |
This is not a quality ranking of the engines. It is a distribution map, and it is deliberately incomplete.
Perplexity has the widest source layer of the three exactly measured engines in the Index overall, so it is not surprising that it is present on every top-ten HR buyer-choice source. Whether it is the widest of all six is not something this release can settle, because three arms carry floors. ChatGPT's four absences are concentrated in the same lightly classified operator surfaces that are thin everywhere — the three collecting-confidence domains plus Reddit.
The two Google rows are blank for a reason worth stating plainly, because the earlier version of this page filled them in. Google AI Overviews recorded nothing at all on 41 of this window's 132 days and Google AI Mode on 11. A whole-window presence flag only accumulates: a missing day can remove a domain from an engine's set and can never add one. Any absence count for those two engines is therefore a count of our own collection gap at least as much as of the engine's behaviour, and there is no way to separate the two from the public release.
The practical consequence is simple: if you only test one engine, you can mistake a source's engine-specific visibility for category authority. The second consequence is the same discipline applied to the measurement itself — if an engine's data has holes, its absences are not evidence.
A vendor whose strongest citations come from LinkedIn, Lever and ClearCompany is in a different position from a vendor whose strongest citations come from Selenios, Umantis and CVViZ. The first set is visible across the measured panel with real evidence behind it. The second set is winning a category shape on a handful of citations.
What this changes for HR software operators
The first HR buyer-choice leaderboard already killed one lazy assumption: the buyer shortlist is not an HR trade-press list.
The engine layer kills a second one: a top-ten source is not automatically a broadly cited source.
For an HR software company, that changes the diagnostic sequence.
First, read the category leaderboard to see what kind of source the answer engines use for the question. In this case, the answer is vendor pages, community platforms and operator surfaces, not editorial publications.
Second, read engine reach before calling any source a durable citation surface. LinkedIn, ClearCompany, US Tech Automations and Lever are cited by all three engines with gap-free collection, and were recorded on every gapped lane as well. Selenios, Umantis and CVViZ are not, and each has fewer than fifteen citations in the whole release.
Third, separate news strategy from buyer-choice strategy. The HR news layer is much more broadly distributed across engines and contains the trade press. The buyer-choice layer is younger and narrower. Treating those as the same source market would erase the only useful signal.
Fourth, before you accept any engine-by-engine comparison — ours or a vendor's — ask what days are missing from each engine's data. An engine that was not collected cannot be shown to have cited nothing.
What to watch next
The next useful read is not another top-ten recap. It is a movement test.
Best-tools, is-it-worth-it and problem-first are each one run date short of the HR evidence floor. When one of them publishes, the question is whether the same narrow sources carry across, or whether the top ten shifts toward broadly cited domains.
Three outcomes would matter.
If Selenios, Umantis, CVViZ or another two-engine domain appears again, the HR buyer layer may be forming around narrow but repeated operator surfaces.
If People Managing People or SelectSoftware Reviews moves into a buyer top ten, the trade press is shape-sensitive rather than absent.
If the next buyer shape shifts toward domains cited by all three exactly measured engines, then this first HR buyer-choice list is a first-published slice, not a durable category pattern.
A fourth is now possible that was not before. Both Google lanes resumed collection on September 25, 2026. Once enough clean days accumulate on them, this same top ten can be read across all six engines, and the answer may look different from either version of this page.
Until then, the clean reading is narrow: HR buyer-choice citations have a first leaderboard, and that leaderboard is more engine-fragmented than the HR news list.
Correction, September 25, 2026
This page originally ranked all six engines by how many of the HR buyer-choice top ten each one missed, put Google AI Overviews at the top of that ranking with 5 of 10, and explained the result with a sentence about AI Overviews being "a search feature with a narrower citation surface". That explanation is withdrawn, and so is the ranking it explained.
The Index's own collection recorded zero citations from Google AI Overviews between August 15 and September 24, 2026, and zero from Google AI Mode between September 14 and September 24 — 41 and 11 days of a 132-day window. Both blackouts were inside the window this page measured, in both the September 21 release it first used and the September 25 release it now uses. The engine that looked narrowest was the engine whose data was most incomplete, and the page presented that as a finding about the engine.
Three figures changed as a result. The headline count of top-ten sources that are not universally cited moved from 6 of 10 to 4 of 10, because the denominator moved from six engines to those with gap-free collection. The news-list comparison moved from 8 of 10 to 10 of 10 on the same basis. The engine table no longer ranks any gapped lane at all.
Correction, September 26, 2026
The September 25 correction below named two gapped engines. There are three: Gemini has 22 intermittent blind days in the same window — May 11 to 15, then scattered single days from August 11 to September 16 — which is why nine sweeps hunting a continuous outage walked past it. Gemini has therefore been removed from every absence count on this page, exactly as the two Google lanes were, and its recorded presences are kept and labelled as floors. Two figures moved again: the news-list comparison from 9 of 10 to 10 of 10, and the buyer-choice absence table from four measurable engines to three. The headline count of 4 of 10 did not move, because all four of those domains are absent from ChatGPT or Claude independently of Gemini. This page's own fourth diagnostic step — ask which days are missing from each engine's data — is the step that found it, applied to us.
Nothing else on the page changed. The segment is identical on both releases — 101 observed runs, 235 cited domains, seven run dates for the buyer-choice shape, and 608 runs across 53 dates for news — and the top ten is rank-for-rank the same, so the leaderboard, the segment rates and the source roles stand as first published.
Method note
Source: the public Machine Relations Index release mri_score_v2.0+2026-09-25+04fcb7fb8f29, generated 2026-09-25, window May 10 through September 25, 2026, with 16,788 observed answer runs, 131,670 citation events and 23,872 cited domains across ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity.
Collection coverage is not uniform across those six engines in this window. The Index recorded no citations from Google AI Overviews on any monitored query between August 15 and September 24, 2026, none from Google AI Mode between September 14 and September 24, and none from Gemini on 22 intermittent days between May 11 and September 16. Every count on this page that could be affected by those gaps is either restricted to the three engines collected without a gap — ChatGPT, Claude and Perplexity — or labelled a floor.
The HR & Talent buyer-choice figures are read from the public HR & Talent / how buyers choose leaderboard. The HR news comparison is read from the HR & Talent / news-driven citations leaderboard. Domain-wide engine reach is read from each domain's public Machine Relations profile under /index/domains/<domain> and means the engine cited that domain at least once anywhere in the full measured panel. It is not a per-segment engine rate, and this article does not infer which engine cited which source inside the HR buyer-choice segment.
FAQ
Do AI engines cite the same HR software sources? Not in a useful sense. In the first published HR buyer-choice top ten, six of ten domains are cited by all three engines the Index collected without a gap in this window; four are not, and the thinnest three of those were recorded on no gapped lane either.
Which engine reaches the whole HR buyer-choice top ten? Perplexity is present on all ten domains in the full release. ChatGPT is absent from four and Claude from one. Gemini, Google AI Mode and Google AI Overviews are not counted, because the Index's collection on all three lanes has missing days inside this window and an absence there cannot be distinguished from a gap.
Why are three of the six engines excluded from the absence counts? The Index recorded no citations from Google AI Overviews for 41 of the window's 132 days, none from Google AI Mode for 11, and none from Gemini on 22 intermittent days. Whole-window presence only accumulates, so missing days can remove a domain from an engine's set but never add one. Presence on those lanes is still real; absence is not measurable.
Does this prove Perplexity cited every source inside HR buyer-choice answers? No. The public release exposes whole-window engine presence by domain, not per-engine rates inside this segment. The claim is about source reach across the release, not per-segment attribution by engine.