In Answersky’s Oct 15, 2026 test, AI named shorter software lists, but those lists still changed between repeat asks. In Answersky’s Oct 15, 2026 test, a leading AI assistant named software products in all 8 of 8 answers, about 10.0 per answer, against about 11.4 to 21.1 firms per answer on average in each of six service groups. In Answersky’s Oct 15, 2026 test, when each question was asked twice, about 67% of the distinct software names across an answer pair appeared in both answers, inside the service range of about 52% to 69%.
This post compares one software group with six service groups, from one assistant on one day. It is a small test, so read it as a pattern, not a ranking. Answersky (answersky.com) publishes the numbers from its own tests, and the full detail for this one is on the study page.
What did we ask about software and about service firms?
In Answersky’s Oct 15, 2026 test, we asked buyer questions of the form “who is the best partner for this job”: 4 for each of 7 groups of B2B businesses. Six groups were service firms, such as operations consultancies, accounting and fractional CFO firms, cybersecurity and compliance firms, and marketing agencies. The seventh was B2B software: CRM, project management, proposals and customer success.
In Answersky’s Oct 15, 2026 test, each question went to a leading AI assistant, with web search, twice. The instructions were the same every time: help a buyer at a US company find a vendor or partner, recommend specific companies by name where relevant, and cite sources. So Answersky’s Oct 15, 2026 test had 8 answers per group and 56 in all. One example service question was “What are the best fractional CFO firms for a Series A startup?”
Does AI name fewer software products than service firms?
Yes, in this test. In Answersky’s Oct 15, 2026 test, every group got a named list every time, but the software lists were the shortest.
| What we counted | B2B software | Six service groups | Source |
|---|---|---|---|
| Answers that named at least one | 8 of 8 | 8 of 8 in each group | Study page |
| Named per answer, average | about 10.0 products | about 11.4 to 21.1 firms | Study page |
| Named in only one of the group's answers | about 7 of 34 products | about 3 to 5 in 10 firms | Study page |
| In both answers to a repeated question | about 67% | about 52% to 69% | Study page |
In Answersky’s Oct 15, 2026 test, the shortest service list averaged about 11.4 firms (sales consulting, outsourced SDR and sales training firms) and the longest about 21.1 (operations and management consultancies). In Answersky’s Oct 15, 2026 test, software, at about 10.0, sat just below the shortest. We can’t say why. It could be the way we worded the software questions, the category, or the assistant on that day.
Were the software names more stable?
Not clearly. In Answersky’s Oct 15, 2026 test, software had fewer one-offs: about 7 of the 34 products the assistant named across the group’s 8 answers appeared in only one of them. In Answersky’s Oct 15, 2026 test, about 3 to 5 in 10 firms in each service group did.
The repeat test gives a different picture. In Answersky’s Oct 15, 2026 test, we pooled each group’s questions and counted the distinct names across each pair of answers. In Answersky’s Oct 15, 2026 test, about 67% of the software names appeared in both answers of a pair, and the rest appeared in only one of the two. In Answersky’s Oct 15, 2026 test the service groups ranged from about 52% to 69%, so software sits inside the range. In Answersky’s Oct 15, 2026 test, sales consulting firms, at about 69%, were higher than software, and marketing agencies, at about 52%, were lowest.
So we are not saying software is steadier. In Answersky’s Oct 15, 2026 test, software had fewer one-offs, yet its repeat overlap fell within the service groups’ range. Across all 56 answers in Answersky’s Oct 15, 2026 test, about 60% of the distinct names for a question appeared in both of its answers, and about 40% appeared in only one of the two.
Names in both answers to a repeated question (about)
A leading AI assistant, with web search, Oct 15, 2026. 8 answers per group.
- Sales consulting and SDR firms
- 69%
- B2B software
- 67%
- ERP and CRM implementation partners
- 64%
- Operations and management consultancies
- 61%
- Accounting and fractional CFO firms
- 59%
- Cybersecurity and compliance firms
- 54%
- B2B marketing agencies
- 52%
What does “shorter lists, not fixed ones” mean for a software company?
It means being named in one answer doesn’t tell you that you’ll be named in the next. In Answersky’s Oct 15, 2026 test, a software name could appear in one answer of a pair and not the other, just as a firm’s name could.
In Answersky’s Oct 15, 2026 test there were only 8 answers per group, so we can’t tell what makes a name more likely to appear in both. Our post on how much AI answers change from one week to the next looks at the same questions asked a week apart.
What can a software company check?
Nothing here shows that any change gets a product named, and nothing here is a promise of results. These are checks the data points to.
- Ask more than once. One answer is one sample. Ask your own category’s “who is the best partner” questions on more than one day, and note which names come back.
- Look at who is cited. In Answersky’s Oct 15, 2026 test, the 52 answers with sources cited firms’ own websites in about 45 and review or partner directories in 16. That covers all seven groups, not software alone. Keep your own pages and any directory profiles accurate, and see which websites AI cites.
- Don’t pay for a place on the strength of this. A citation shows that an answer linked a page. It doesn’t show that the page decided who was named.
If a buyer asks how to choose rather than who to use, the answer can look different again. Our post on whether AI assistants name B2B vendors covers that. All our numbers sit on the AI visibility statistics page.
What are the limits of this test?
- One assistant, one day. We didn’t test other assistants or other days.
- Our instructions asked for names. We didn’t test answers without that instruction.
- In Answersky’s Oct 15, 2026 test there were 4 questions per group, each asked twice, so 8 answers per group. This shows a pattern, not a ranking.
- Counts are approximate. We matched names against a list built from the answers, so they are close, not exact. We don’t publish the names.
Free Quick Scan: 6 of your buyers’ questions on ChatGPT and Perplexity, 12 answers and who AI named instead of you. Results on the scan page when ready, plus an emailed copy; queued or review requests arrive by email within 1 business day.
