Gusto was the product ChatGPT and Gemini recommended most, named in 5 of the 6 answers they gave when we asked how to choose payroll software on 3 August 2026. It finished one answer clear of the next two, ADP and OnPay, at 4 of 6 each.
We put six buyer questions to the two assistants named in this run, ChatGPT and Gemini, gave each question one live web search, and recorded every product they named and every source page behind the answers. We did not test any other assistant, so across all 6 questions this page speaks only for what those two said on the day, not for the wider market. Everything below comes from that one measurement.
The products they named
Gusto led at 5 of 6 answers (83%). Two tools tied for second at 4 of 6 (67%): ADP and OnPay. Paychex and SurePayroll each landed 3 of 6 (50%). That top five is where the recommendations pooled; every other product was named in 2 answers or fewer.
Four familiar names sat at 2 of 6 (33%): Patriot Software, QuickBooks Payroll, Rippling and TimeTrex. Below them stretched a long tail of 12 products each named exactly once (1 of 6, 17%), from Justworks and Zoho Payroll to Payroll4Free.com and eSmart Paycheck.
Two brands appeared under more than one label, and we kept them separate because the measurement did. ADP was named as "ADP" in 4 answers and as "RUN Powered by ADP" in 1 more. QuickBooks showed up three ways: "QuickBooks Payroll" (2), "QuickBooks" (1) and "QuickBooks Workforce" (1). Patriot came up as "Patriot Software" (2) and "Patriot Payroll" (1). We are not adding those together, because the assistants said them as separate names, and the exact wording is part of what we measure.
The shortlist changed with the question
This is the section to stare at. The six questions read almost the same in plain English, but the answers did not. The plain "best payroll software" question returned 7 names: Gusto, RUN Powered by ADP, Patriot Software, Rippling, Justworks, OnPay and Paychex. Ask instead for the "best free payroll software" and 6 of those 7 vanish. The free answer was Payroll4Free.com, ExcelPayroll, HR.my, HelloPaid, TimeTrex, eSmart Paycheck and Gusto, with only Gusto carrying over from the plain "best" list.
"Free" and "affordable" sound like neighbors, yet they pulled completely separate rosters with zero products in common. "Best free payroll software" gave the 7 names above. "Most affordable payroll software" gave just 3: ADP, Zoho Payroll and SurePayroll. Not one product sat on both lists.
The "most affordable" question was the narrowest of the six, 3 products against a high of 8 for both "small teams" and "what should I use." It was also the only question that did not name Gusto, the overall leader, and the only place Zoho Payroll appeared at all.
Twelve products were each locked to a single question out of six. "Best free" alone introduced five that appear nowhere else in the test: Payroll4Free.com, ExcelPayroll, HR.my, HelloPaid and eSmart Paycheck. "Small teams" was the only home for Homebase, Patriot Payroll and QuickBooks Workforce. "What should I use" was the only place plain QuickBooks surfaced, "best" the only place Justworks and RUN Powered by ADP appeared, and "most affordable" the only place Zoho Payroll did. A buyer who only typed "best" would never learn those eight had entered the conversation.
The point holds across all 6 questions: there is no one AI answer for a category. The word a buyer picks, "best," "free," "affordable," "for small teams," decides which shortlist they see, and neighboring words returned lists that barely touched.
Where the answers came from
Reddit fed more answers than any other source, cited on all 6 of 6 questions (100%). Forbes came next at 5 of 6. Then the interesting one: gusto.com, a vendor's own site, also at 5 of 6.
The vendor-domain pattern is the real finding, and at least one company's own site was cited in every one of the 6 answers. The owned domains the assistants leaned on line up with the brands that ranked: gusto.com cited on 5 questions and Gusto named on 5, adp.com cited on 4 and ADP named on 4, patriotsoftware.com cited on 4 with Patriot named as Patriot Software on 2 and Patriot Payroll on 1, quickbooks.intuit.com cited on 3 with QuickBooks named under three labels across 3 answers. Even the free list did it: payroll4free.com was cited on the one "free" question and Payroll4Free.com was named in that same answer. When a company's own pages sit in the sources an assistant reads, that company tends to land in the answer.
There is one clean exception worth flagging: OnPay was named in 4 of 6 answers with no owned domain cited anywhere in the run. Paychex (3 of 6) and SurePayroll (3 of 6) did the same. So a strong showing on the third-party pages, Reddit at 6 of 6 and Forbes at 5 of 6, can carry a product even when its own site is never pulled.
The rest of the map is community threads and editorial roundups: YouTube on 3 questions, CNBC on 2, TechnologyAdvice on 2, a personal blog at karllhughes.com on 2, and single cites from Quora, Indeed, U.S. News, Helcim, Springly, MyShortlister and equalpaytoday.org. It is forum posts, big-media lists and individual writers, not an official register.
The review directories were nearly absent. G2 was cited exactly once, on the plain "best" question, and Capterra did not appear at all across the 6 questions. In a category where vendors spend hard on their G2 and Capterra profiles, neither assistant reached for those directories to answer these buyers.
If you sell payroll software
The lesson is narrow and useful: to get named, be in these exact sources, because the source list, not your market share, decides the answer. Every one of the top four brands either had its own domain in the citations or a heavy presence in Reddit and Forbes, and usually both.
From this data, that means three moves. Be in the Reddit threads that answer these questions, cited on 6 of 6. Get into the Forbes-tier roundups, cited on 5 of 6. And keep your own site directly answering the buyer's question, since owned domains fed all 6 answers. OnPay's 4 of 6 with zero self-citations shows the third-party route works on its own, and G2 (cited once) and Capterra (cited zero times) show where the effort is wasted for these two assistants.
The full list, counted
| Product | Named in | Share |
|---|---|---|
| Gusto | 5 of 6 | 83% |
| ADP | 4 of 6 | 67% |
| OnPay | 4 of 6 | 67% |
| Paychex | 3 of 6 | 50% |
| SurePayroll | 3 of 6 | 50% |
| Patriot Software | 2 of 6 | 33% |
| QuickBooks Payroll | 2 of 6 | 33% |
| Rippling | 2 of 6 | 33% |
| TimeTrex | 2 of 6 | 33% |
ADP also appears once more as "RUN Powered by ADP," and QuickBooks Payroll shares its brand with single cites for "QuickBooks" and "QuickBooks Workforce." We list the labels as the assistants said them rather than merge them into counts the data never showed.
Why we ran this
Glotier does not sell payroll software, so we are correctly nowhere in these answers, and we will not pretend otherwise. We chose payroll for this week precisely because we have no stake in it, which is the honest way to show what the check does.
This run is exactly what we do for a customer, aimed at their own category instead: six buyer questions, the real assistants, every product and every source written down, so you can see whether you are named, who is ahead of you, and which pages you would have to be on to move. If you sell in a category, you can run the same check yourself, free, no account, in about a minute.