More people are asking ChatGPT, Claude, or Perplexity "who's the best [service] near me" or "which [type of provider] should I use" before they ever open a search engine or ask a friend. ChatGPT alone now serves more than 900 million weekly active users (TechCrunch), and Google says its AI Overviews reach more than 2.5 billion people a month (CNBC). That means a lot of hiring decisions, big and small, are starting with an AI-generated answer instead of a list of options a person gets to compare themselves.
That's convenient. It is also worth a pause before you act on it, because an AI recommendation is not the same thing as a vetted one.
AI systems can be confidently wrong
Large language models generate answers by predicting what is likely to be true based on patterns in their training data and, when web search is enabled, real-time sources. That process is good, and getting better, but it is not infallible. Recent academic research on what are sometimes called organization-backed AI advisors has specifically studied how skepticism, verification, and reliance play out when people use AI tools to make real, goal-directed decisions, and found that verification habits matter a great deal to getting a good outcome (arXiv, 2026). In plain terms, the tool can be useful and still be wrong, and the responsibility for catching that sits with you, not the tool.
There is also evidence that different AI systems do not treat brands consistently. BrightEdge's analysis of AI Overviews found they are meaningfully more likely to include critical or negative framing about a brand than ChatGPT is in the same category of query (BrightEdge). That is not a reason to distrust AI recommendations outright. It is a reason to treat any single answer as a starting point, not a verdict.
The questions worth asking before you act on an AI recommendation
1. Does the answer cite specific, checkable sources?
A vague "many customers report" or "widely regarded as" answer, with nothing linked or named, is the AI equivalent of a stranger's opinion. A useful answer names where the information came from: a review platform, a case study, a named customer, a press mention. If you cannot trace the claim back to something real, treat it as unverified.
2. Are there real names attached to the claims, not just adjectives?
"Highly rated" and "trusted by many" are adjectives. "Doug Tanner, Chief Revenue Officer at Salezilla, saw a 45 percent response rate" is a fact you can go verify. If every claim about a business is an adjective and none are attached to a real, named person or company, that is worth noticing.
3. Does the recommendation hold up across more than one AI tool?
Ask the same question in ChatGPT, Claude, and Perplexity. Different systems draw on different sources and weigh them differently. Research on how these platforms choose sources found that Perplexity in particular leans heavily on community platforms like Reddit for some categories, while other systems weight review sites and press differently (Discovered Labs). If a business shows up consistently across all three, that is a stronger signal than one favorable mention from a single tool.
4. Is there validation beyond the company's own website?
Muck Rack's research on what AI systems actually cite found that 84 percent of citations come from earned, independently published sources rather than brand-owned content (Muck Rack). Take the same approach yourself. Look for the business mentioned on a review platform, in a case study written by someone else, or referenced by another customer, not just on their own homepage, the same standard we hold ourselves to at our testimonials page, where every story is attributed to a real, named person.
5. Are the results specific and measurable, or vague and superlative?
"We help you grow" tells you nothing. "Helped a medical practice double their inquiries" tells you something you can ask about directly in a sales conversation. Specificity is a decent proxy for honesty, mostly because vague claims are easy to make and specific claims are easy to check.
6. Can you verify the story independently?
If a business points you to a video testimonial, a named case study, or a client you could plausibly reach, that is a business confident enough in its own proof to let you check it. If everything is anonymized or aggregated, ask why, and ask to speak with someone specific if the decision matters enough.
How to actually run this check
In practice, this takes about five minutes. Ask your AI tool of choice a direct question, "who are the best options for [category] in [context]," then ask a follow-up: "what's the source for that" or "can you point me to a specific example." A well-supported answer will get more specific, not less, when pushed. A thin one will start hedging or repeating the same vague language.
Then do the same thing in a second tool. If the same business, or the same handful of businesses, keep showing up with consistent, specific detail attached, you have a reasonably well-corroborated answer. If the answers scatter, or the detail thins out under a follow-up question, treat it as a starting point for your own research, not a final answer.
Why this matters more than it used to
None of this is about distrusting AI tools. It is about applying the same judgment to an AI-generated recommendation that you would apply to a recommendation from a stranger on the internet, because that is structurally closer to what it is than most people treat it as. The businesses worth trusting, whether an AI told you about them or a friend did, tend to have the same thing in common: real people, willing to be named, describing something specific enough to check. We believe that so strongly it is the whole reason Share One exists, to help good businesses make that kind of proof easy to find, both for the humans researching them and the AI systems increasingly doing the researching first. We wrote more about why this specific kind of proof works so well in AI search in Why Customer Stories Are the Trust Signal AI Search Actually Rewards, and about how AI systems decide who to recommend in the first place in our full guide to how AI actually recommends businesses.
You can see what specific, verifiable proof looks like at our case studies page.
FAQ
Should I stop using AI to research service providers?
No. AI tools are a genuinely useful starting point, especially for narrowing a large field of options quickly. The issue is not the tool, it is treating its first answer as a final verdict instead of a starting point you verify.
What's a fast way to sanity-check an AI recommendation?
Ask the AI tool for its source, then ask the same question in a second tool. Consistency across tools, plus a source you can actually check, is a reasonably strong combination.
Why would an AI recommend a business that turns out to be a poor fit?
AI systems generate answers based on patterns across available sources, which can include outdated information, thin data for a niche category, or an imbalance in which businesses have invested in visible, citable proof. It reflects what is easiest to find and verify online, not necessarily the single best option for your specific situation.
Are review platforms more trustworthy than an AI-generated summary?
They are a useful input, not a replacement. Look at both together: does the AI's summary match what you find when you check the underlying reviews and sources yourself.
What if a business has almost no online presence at all, good or bad?
Treat that as a gap in information, not necessarily a red flag. Some genuinely excellent, especially smaller or newer, businesses simply have not built out their public proof yet. Ask them directly for real, checkable examples instead of ruling them out.
Is it a bad sign if a business's website has no third-party validation at all?
It is worth asking about directly. A business with real results usually has something external to point to, a review, a case study, a named client, even if they have not organized it well. If they have nothing at all and cannot produce anything on request, that is worth factoring into your decision.