Picture someone asking ChatGPT for the best budgeting app in Australia. Or the best health insurer, or somewhere to stay in Byron. They get a handful of names back and one clear recommendation. Your brand might make the list. The odds are it won’t be the one AI tells them to choose.
AI knows a lot of brands. But it recommends very few of them. LEOPRD’s Reputation to Revenue 2026 report, which launched this week, tracked 27 brands across eight AI platforms, and found that when people ask AI for a recommendation without naming a brand, seven times out of ten the brand isn't AI's first choice. Half of those times, it isn't mentioned at all.
That matters, because people used AI around 20 times more before buying something than after it. They’re researching, comparing and deciding, often before a brand knows they’re looking.
Being overlooked costs money. Brands lose sales they never knew they were in the running for. And when AI gets the details wrong, they pay again in returns, complaints and support calls.
When we released our first Reputation to Revenue report in April 2025, the big question was whether AI could find your brand. That is still important. But AI has evolved rapidly in 18 months, and the question brands need to answer now is why AI names them, then recommends someone else.
“Being found isn’t the same as being selected. If ChatGPT mentions you alongside four competitors and then tells the customer somebody else is the better choice, you’ve achieved visibility, but you’re unlikely to be chosen.”
Celia Harding, founder, LEOPRD
Missing, Mentioned or Most Wanted is LEOPRD’s way of measuring how a brand shows up in AI answers: left out altogether, named but passed over, or recommended first.
TL;DR (Key findings)
- Brands were missing from 36% of AI answers where no brand was named, mentioned but passed over in 35%, and AI’s first choice in just 29%.
- LEOPRD’s modelling links being left out of AI answers to around A$2.9m in potential exposure for every A$100m of Australian online retail spending.
- People used AI around 20 times more before a purchase than after it: 27% of prompts were research, comparison and checking, compared with 1.3% for help after buying.
- 86% of the sources behind AI answers came from outside brands’ own websites and channels.
- Only 40% of repeated tests gave the same recommendation every time.
- Read the full findings in the Reputation to Revenue 2026 report.
Why does ChatGPT recommend my competitor instead of me?
When someone asks AI an open question, like “What’s the best budgeting app in Australia?”, it builds a shortlist and then picks a favourite. Most brands make the shortlist some of the time. Far fewer get recommended.
A high visibility score can hide this. Fintech brands in the study appeared in 92% of relevant answers, but were AI’s primary recommendation in only 32% of the answers they appeared in.
AI weighs up the evidence it can find about each brand, then recommends the one that’s easiest to justify for that customer. It’s about what the evidence says about you, not how often your name comes up.
What does it cost to be left out of AI answers?
When people asked AI for a recommendation without naming a brand, the average brand in the study was left out of 36% of answers. LEOPRD’s modelling links that to around A$2.9m in potential exposure for every A$100m of Australian online retail spending, and around £3.1m for every £100m in the UK. These are modelled estimates, not measured revenue losses.
Being misunderstood costs money too. When AI gives a customer incomplete or wrong information, they can end up with the wrong product or service. The business then carries the impact for it in returns, complaints, support calls and customers who leave.
“A marketing team can optimise content, but it can’t optimise away a product problem, repeated customer complaints or an absence of independent evidence.”
Celia Harding
Where does AI get its information about my brand?
Mostly from other people. 86% of the sources AI cited came from outside brands’ own websites, including media coverage, reviews, comparison sites and online communities.
Comparison sites were the most common source, making up around 28% of what AI cited. And when people asked AI what could go wrong with a brand, reviews jumped from 11% to 33% of its sources.
The mix changes by category. In travel, 41% of the sources behind AI’s first-choice recommendations were editorial coverage. B2B software leaned more on comparison sites, while health and care relied more on brands’ own information and research.
For anyone in PR, this is familiar ground. Shaping what credible third parties say about a brand is what good communications teams already do, and that work shapes what AI says too.
Why is one AI answer never enough?
AI reputation is the pattern of how AI platforms describe, compare and recommend a brand across many questions, platforms and points in time.
One AI answer tells you very little. Only 40% of repeated tests gave the same recommendation every time, even when LEOPRD asked the same question, on the same platform, in the same market.
“Typing your company into ChatGPT and screenshotting the answer is not reputation monitoring. One answer is an observation. The real task is to understand which issues, sources and narratives recur across platforms and over time.”
Celia Harding
How can your brand become AI’s first choice?
- Pick the answers that matter. Start with the questions customers ask AI just before they buy.
- Work out which problem you have. A brand that never appears needs presence in credible sources. A brand that appears but isn’t endorsed needs better evidence of why it should be.
- Fix the evidence behind the answer. To change the answer, change what your website, media coverage, reviews and comparison sites say about you. But PR and SEO can't fix a poor customer experience. If reviews keep raising the same problem, fix the problem first.
- Get the whole business working from the same plan. AI doesn't see your org chart. What it finds about you comes from every team, from product, customer service and legal to HR, marketing and comms. If that information is patchy or inconsistent, AI finds it harder to judge you clearly.
- Test again, and connect it to the business. Ask the same questions over time and track whether the results move alongside leads, complaints and churn.
Missing, Mentioned or Most Wanted: how brands show up in AI answers
Later this week, we’ll look at how Australians use AI to check their buying decisions. More findings will be presented at humAIn in Sydney on 13 and 14 October.
Frequently asked questions
What is LEOPRD’s Reputation to Revenue 2026 report?
Reputation to Revenue 2026 is LEOPRD’s research into how AI platforms describe and recommend brands, and what that means commercially. It analysed 31,200 AI answers across 27 brands, seven categories and eight AI platforms in Australia and Great Britain, plus 420,903 anonymised prompts from Prompt Cowboy covering Australia, the UK and the US.
How often do brands become AI’s primary recommendation?
Brands were AI’s first choice in just 29% of answers where the customer didn’t name a brand. They were mentioned but passed over in 35%, and missing altogether in 36%. That means brands missed out on first choice in around seven out of ten answers.
What does it cost a brand to be left out of AI answers?
LEOPRD’s modelling links being left out of AI answers to around A$2.9m in potential exposure for every A$100m of Australian online retail spending, and around £3.1m for every £100m in the UK. These are modelled estimates, not measured revenue losses.
Where does AI get its information about brands?
86% of the sources AI cited came from outside brands’ own websites, including media coverage, reviews, comparison sites, communities and research. The mix changes by category: travel leans on editorial coverage, while B2B software leans on comparison sites.
Why does AI give different answers to the same question?
AI answers change between platforms, questions and points in time. LEOPRD found only 40% of repeated tests gave the same recommendation every time, so a brand’s AI reputation needs to be tracked as a pattern rather than judged from one answer.
The bottom line
AI knows most brands, but it recommends very few of them. With seven in ten answers going to someone else, being mentioned isn’t enough to win the sale. The brands AI picks first are the ones with clear, credible evidence about them in the places AI looks. That evidence can be built, measured and improved.
Find out where your brand stands
Download the Reputation to Revenue 2026 report to see how AI chooses which brands to recommend, and what it costs to be overlooked.
Or book an AI Visibility Audit to find out whether ChatGPT, Gemini, Perplexity and Copilot have your brand missing, mentioned or most wanted, and what it would take to change the answer.
About the author
Celia Harding is the founder of LEOPRD, an AI visibility and reputation advisory specialising in Language Engine Optimisation (LEO). LEOPRD helps brands understand and influence how they show up in AI platforms including ChatGPT, Google Gemini and Perplexity. Celia is the winner of CommsCon PR Professional of the Year and B&T Women Leading Tech PR 2026, and LEOPRD was named Best New Agency in Asia Pacific at the 2026 SABRE Awards.
About the research
LEOPRD’s Reputation to Revenue 2026 report combines consumer behaviour data, AI platform analysis and interviews with communications, marketing and behavioural science leaders. LEOPRD analysed a random, anonymised and de-identified sample of 420,903 prompts from Prompt Cowboy’s dataset of 10.4 million prompts, collected between November 2025 and August 2026 across Australia, the UK and the US. No full prompt text or personal information was included. LEOPRD separately analysed a sample of 31,200 AI responses from a monitoring programme of more than 214,000 responses, covering 27 brands, seven categories, eight AI platforms and two markets (Australia and Great Britain). No current LEOPRD clients were included in the study.
Sources: LEOPRD, Reputation to Revenue 2026: The Cost of Being Overlooked. Commercial figures are LEOPRD-modelled exposure estimates, not measured revenue losses.