Check first, because the names AI gives change from one ask to the next: in Answersky’s Oct 15, 2026 test of one AI assistant with web search, only about 60% of the firms it named for a B2B buyer’s question appeared in both answers when we asked twice. We don’t name that assistant, so this isn’t a measure of ChatGPT in particular. Then choose who does the fixing: you, an agency or consultant, or a service that does it for you. Answersky offers two paid plans for that, Guided and Done for you, and our free Quick Scan shows where you stand before you spend anything.
We’re Answersky (answersky.com), and we sell one of the options below, so we say what it costs and what we can’t promise. Every figure here comes from our own tests, with the date and the method beside it.
Why does AI recommend my competitor over me?
We can’t see why an assistant picks the names it does, only the answers it gives. Four things are worth knowing, and the last one is a suggestion, not a finding.
1. The list isn’t fixed. In Answersky’s Oct 15, 2026 test, a leading AI assistant, with web search, named firms in all 56 answers to B2B buyer questions, about 14 per answer. We asked each question twice. Counting each firm once per question across all 28 questions, about 60% of the firms named for a question appeared in both of its answers.
Same firms on a repeat ask
Answersky's Oct 15, 2026 test. Share of the firms named for a question that appeared in both of its two answers, pooled by group.
- Sales consulting and outsourced sales
- 69%
- B2B software (for comparison)
- 67%
- ERP and CRM implementation partners
- 64%
- Operations and management consultancies
- 61%
- Startup accounting and fractional CFO firms
- 59%
- Cybersecurity and compliance firms
- 54%
- B2B marketing agencies
- 52%
So a competitor you saw in one answer may not be in the next, and a single check where you weren’t named doesn’t show you are never named. It also means being named once isn’t a standing.
2. The answers link several kinds of page. In Answersky’s Oct 15, 2026 test, the assistant searched the web in 52 of the 56 answers and linked sources in all 52. It cited other websites (articles, media, blogs and lists) in 49 of them, a named firm’s own website in about 45 and review or partner directories in 16. An answer can cite several types, so these counts overlap. They show that the assistant looked beyond firms’ own websites, not why one firm was named and another wasn’t. A citation shows a page was linked, not that the page decided who was named.
3. A site that blocks OpenAI’s search crawler won’t be shown in ChatGPT search answers. OpenAI’s crawler documentation says a site that is opted out of OAI-SearchBot isn’t shown in ChatGPT search answers, though it can still appear as a navigational link. Letting the crawler in doesn’t ensure a place in an answer or a recommendation. If your robots.txt or your host’s bot setting blocks it, that is worth fixing first.
4. Your pages may not say what buyers are told to check. This one is a suggestion, not a finding. In Answersky’s Oct 12, 2026 test of accounting firms, the assistants told buyers to check experience with clients like you in 26 of the 27 “how do I choose” answers, and fees and billing in 24 of 27. The counts are in the “What were buyers told to check?” table on our accounting firms study page, and a second AI model extracted the checks, so they are approximate. That is advice the assistants gave buyers. It doesn’t show that changing a page makes an assistant recommend you. But a buyer who follows that advice will look for who you serve and how you charge, so check that your pages say it plainly.
We’re in the same position. On Sep 30, 2026, our first scan of answersky.com asked ChatGPT, Gemini and Perplexity the questions our buyers ask, each in three wordings, and none of the answers named us. ChatGPT and Perplexity cited 15 sites in those answers, and answersky.com wasn’t one of them. Our scan log shows what we changed since and what moved.
What should I check first?
Find out whether you are named at all, and who is named instead. You can do it by hand: ask the questions your buyers ask on each assistant with web search on, write down who is named and which sites are linked, then ask again another day. How to check if ChatGPT mentions your business walks through it.
Or use our free Quick Scan: 6 of your buyers’ questions on ChatGPT and Perplexity, 12 answers, and who AI named instead of you. It shows named or not in each answer. It has no score and no fixes. Start a free Quick Scan. Results on the scan page when ready, plus an emailed copy; queued or review requests arrive by email within 1 business day.
One ask is one sample. Because the lists change, read the first result as a snapshot, then check again before you decide what to pay for.
Who can help: do it yourself, an agency or a done-for-you service?
All three start from the same checks, and they differ in who does the work after them.
| Option | Who does the work | It can fit if | What to ask or check |
|---|---|---|---|
| Do it yourself | You, or someone on your team | You have someone to ask the questions, record the answers and edit your pages, and the checks above are enough | Keep the questions fixed, ask more than once and write down the dates |
| An agency or consultant | The people you hire, with your input | You want a person to own the job and have a budget for it | What they will change, whether they show you the real dated answers, and whether they promise you will be named |
| A done-for-you service | The service, with your approval | You want the changes made and only need to approve them | The same three things to ask |
Whoever you pick, ask to see the recorded answers, with dates, before and after the work. Since in Answersky’s Oct 15, 2026 test only about 60% of named firms repeated, one screenshot of “we were named” proves little. And nobody controls what an assistant says, so be careful of anyone who promises you will be named.
Answersky is one option, with two plans that include page fixes:
- Guided is $99 a month. We write the page fixes and article outlines, and you paste them in. We also name the sites to get listed on. A weekly check on ChatGPT, Gemini, Perplexity and Claude is included.
- Done for you is $549 a month. We make the changes on your site and on the lists AI cites, and you approve each one before it goes live.
Both can be cancelled any time. We can’t say when, or whether, a change will lead to you being named: each month the report shows what the answers say and what moved. What we do, what we don’t do and the full plans. There are other providers, and we haven’t compared their prices here.
How will I know it worked?
Ask the same questions again, on a schedule, and compare the answers over several checks, not one. That is the method behind our scan log. The number to watch is how often you are named across repeated asks, because a single answer can swing either way.
What are the limits of our test?
All 56 answers in Answersky’s Oct 15, 2026 test came from one AI assistant, with web search, on one day. It was told to recommend firms by name where relevant, and we didn’t test answers without that instruction. Each of the 7 groups had 4 questions, each asked twice. It shows a pattern, not a ranking, and the consumer apps can answer differently from the APIs we use. The full test page lists the limits and the counting method.
Nothing here says any change will get your company named.
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.
