Unfortunately, it’s not enough to be simply mentioned by ChatGPT anymore. Especially when your competitor is the star of the answer. If your brand appears in a conversation, it feels like a win, but a recommendation is what makes all the difference.

Imagine someone asking AI to recommend a hospital or a consultant. Your institution appears in the response, but another company gets named first, with a more convincing explanation of why someone should choose them. Your brand is dwarfed by the competition, and there’s no reason someone will go for anything other than the first option, especially when it comes to their health. Although people want the first choice wherever.

To figure out what’s happening, Kooba recently ran an AI visibility test across very different sectors, including retail, engineering, education and payroll software. We looked at thousands of AI-generated answers, looking at how frequently brands appeared, where they featured in responses and which sources were being used. Different businesses, different audiences, different competitors. One common theme…

AI knows the brand but doesn’t recommend it. Aptly named the AI recommendation gap, this highlights a subtle yet all-important difference in how brands should be measuring their AI visibility.


Why isn’t visibility the whole story?

For years, businesses have worked to get their websites ranking as highly as possible on Google. A top result is literally a top result in traditional SEO, and the thinking carried over. Now, AI platforms are adding another consideration: not just whether your business appears in an answer, but how it's presented alongside the competition.

We saw two different scenarios in our research: brands were mentioned loads, rarely as the first option. Others were mentioned fewer times but led the recommendations when they were. Quantity vs. quality of appearance.

  • Infrastructure services: One organisation for a particular service topic was mentioned most, named in 23% of all answers, while it was first in just 1% of those answers.
  • Consumer retail: A kitchen retailer was named more than any competitor in 47% of all answers about local showrooms, while it was first in only 5% of those answers.
  • Professional education: Two competing providers were named a similar amount while one was named first in 55% of all answers, compared with 23% for the other.

Your competitors getting more prominent placements matters more than just being visible…

A 2026 study from Parse, examining almost 744,000 brand mentions across ChatGPT Search and Google AI Mode, found that only 12.9% of those mentions occupied the first position in a ranked recommendation.

Of course, the first brand mentioned isn't always the strongest recommendation, and neither metric proves that a customer will make a purchase. If a potential customer is asking for the best option, appearing –somewhere– in a lengthy answer is like being the wallpaper. What needs to happen for you to become the focus?


About the data:

Snapshots: 5 organisations across 5 sectors (infrastructure services, homeware retail, professional education, building services engineering, automation software)

Platforms tracked: ChatGPT, Perplexity and Google AI Overviews

AI answers analysed: Roughly 6,000.


Does AI know why someone should choose you?

When we looked at what AI engines retrieved and cited, the chatbots tasked with answering queries looked for the sites that (understandably) best answered the question. They were specific with details and backed by evidence, not broad about sector-fluid statements without much to support them.

Specialist service pages with clear, tangible information about actual projects performed well in several of our snapshots. One organisation's technical assessment pages were cited on almost every retrieval. In other words, the engine referenced the source in its answer (cited it) almost every time it pulled up the page while researching (retrieved it). Elsewhere, a practical guide to payroll audit readiness generated more than two citations per retrieval, on average.

Performing a lot less well were the broader thought-leadership articles, generic service descriptions and location pages we examined. It becomes clear that it’s not necessarily about how much content you've written. It's about what that content helps someone understand.

Here are two ways to describe the same business.

We're a leading engineering consultancy delivering innovative solutions for clients across multiple sectors.

Or:

We provide overheating assessments for residential developments, including TM59 analysis, with examples of completed projects and the stages involved.

One sounds like countless other engineering businesses, while the other matches specifics around what a user might ask.

Recognising a business and finding evidence relevant to a particular question are not the same task.

An October 2026 research paper, examining more than 158,000 AI-generated answers, found that pages closely matching the questions AI systems searched for were more likely to be cited. Retrieval patterns explained considerably more of the variation in citations than individual page features.

The researchers also found that AI could mention brands based on existing model knowledge, independently of the pages it retrieved.

In other words, an established reputation might help your business enter the conversation. But if someone asks a more specific question about your expertise, suitability or offering, the available evidence becomes important.


What happens when your website doesn’t answer a question and someone else does?

The next findings become particularly interesting for content strategy.

In four of our five tests, the most prominent cited source was a third-party directory, buyer's guide or ranking site that didn't feature the brand we were analysing. These were sources helping inform answers to questions potential customers were asking. Imagine spending years building a reputation in your industry, only for a potential customer to ask AI for the best providers and receive an answer informed by a third-party list you're not on. It's an undesirable position only fixed by going beyond updating your homepage.

AI search doesn't exclusively rely on your website to build a picture of your business. Depending on the question and platform, it can draw on industry publications, comparison sites, directories, independent reviews and other external sources.

This creates two opportunities:

First, what your website says about you. Are you offering clear, relevant information that supports your expertise and answers the questions customers actually ask?

Second, what the wider web says about you. Are you accurately represented on the platforms, publications and directories customers and AI systems use to compare businesses?

And there's no universal answer about which source matters most. A 2026 analysis by DeltaV Digital found that preferred citation sources varied considerably by industry. Listicles accounted for 61% of citations in its B2B technology services sample, while university programme pages dominated its higher-education sample.

Another interesting point to note was that only 35.7% of the named or mentioned brands were supported with a cited source. The rest the AI recalled based on memory and existing knowledge of them.

Is your best content in the wrong place?

We also noticed another, more practical issue across the findings.

Some businesses had useful content, but it wasn't necessarily presented in a way that made it easy to find, navigate or reference. In two cases, important service information was primarily available through brochures or specification PDFs. These examples performed poorly in the monitored AI answers.

This doesn't mean PDFs are invisible to AI. Google explicitly supports indexing PDF content, and other search systems can retrieve documents too.

However, a dedicated HTML page gives you an opportunity to make your offering clear and accessible within the website's navigation, link to relevant case studies and keep information up to date. These are similar principles that support inclusive design. Our specialist accessibility agency AccessPoint touches on the structure aspect, outlining how PDFs Aren’t As Accessible As You Might Think, which applies to AI access too.

Rather than immediately producing more content or replacing old pages completely, take a closer look at what's already there: Does it have enough detail, answer the customer’s question, and would anyone understand it if they weren’t familiar with your business?

How to close the recommendation gap?

We’ve covered that the answer isn't publishing hundreds of articles designed exclusively for AI. Much of what matters is familiar territory: good content, credible evidence, clear website architecture and a proper understanding of your audience.

Kooba’s AI visibility data gives us more detail in identifying where those things might be falling short.

1. We find the questions where you're mentioned but not leading

By looking beyond your overall AI visibility percentage.

We unearth which questions matter to your specific industry and where your competitors are showing up. That’s surface level, but we can go a lot deeper and tailor solutions to different problems. For example, a low mention rate could indicate limited relevance. Frequent mentions but a poor position could suggest that other businesses are being presented more favourably.

We look at qualitative, actual responses on top of quantitative dashboard scores.

2. We work out why customers would have a reason to choose you

Put your consumer hat on and see if trusted partner, industry-leading or innovative solutions mean anything without supporting examples.

What matters to them is named clients, relevant projects, service processes, specialist expertise, delivery timelines, qualifications and clear pricing information to set your business apart.

For some businesses, that might involve developing a dedicated service page. For others, improving existing case studies or guides explaining how a particular offering works could be more valuable.

3. Don't leave comparison questions entirely to competitors

Buyers ask questions like:

Which provider is best suited to a company of our size?

What's the difference between these two services?

What are the alternatives to this software?

These feel close to the thinking behind a purchasing decision. Your content should help people understand situations and how different solutions might be more appropriate.

No need to publish an aggressive competitor comparison for every rival. Practical buying guides, transparent service explanations and honest comparisons between approaches can be just as useful.

4. We look beyond your own website

Identify the external sources appearing repeatedly in answers about your industry.

Are your competitors being mentioned by publications, directories or review platforms where your brand is absent?

Are your existing listings accurate and up to date?

Kooba can point to opportunities for editorial coverage, industry partnerships, credible directory inclusion or better third-party information about your services. The goal is a consistent, evidence-backed presence wherever customers are likely to research their options.

5. We keep measuring what changes

AI recommendations aren't a fixed set of search rankings.

They can vary between platforms, across different questions and even between repeated answers to the same question.

A single visibility snapshot is useful for identifying opportunities. Regular monitoring tells you whether the changes you've made are working over time. We cover this in more detail in our blog about How Regular AI Visibility Tracking Provides a Share of Voice.

Like with this research, Kooba tracks mention frequency, prominent placements, citations and how your brand is described. Where possible, connect these measures with real business outcomes, including relevant enquiries and conversions.

Being known is only half the job

Generative Engine Optimisation (GEO) is the process of improving your website so that AI search tools can better understand, interpret and reference your content, as defined in How Your Website Can Win AI Search and Traffic. And as we’ve seen here, it’s about recommendation on top of visibility.

The biggest takeaway from our research isn't that businesses need to produce an entirely new type of content for AI. It's that many established brands already have the expertise, projects, services and reputation customers are looking for. They just aren’t always communicating it clearly enough, in the places where people are making comparisons.

And while AI search has introduced new considerations, there's something reassuringly familiar about the solution. Understand what your customers need to know. Make the answers easy to find. Give them credible reasons to trust you. And make sure the wider web reflects the business you've built. Basic stuff with a new lens.

Go so much further than appearing in AI-generated answers and make sure those answers are doing your business justice.

At Kooba, we use ongoing GEO analysis to identify where brands appear in AI search, how they compare with competitors and which content or external sources offer opportunities for improvement.

If your business is regularly being mentioned but rarely placed first, we will help you understand why.



Frequently asked questions

What is the AI recommendation gap?

The AI recommendation gap describes the difference between how often a brand appears in AI-generated answers and how often it occupies a leading recommendation position. It helps businesses distinguish between general visibility and prominence in commercially important responses.

How can a business measure AI recommendations?

AI visibility platforms can track brand mentions, position, sentiment and cited sources across a defined set of prompts. These metrics should be reviewed together, with explicit recommendations distinguished from simple mentions, and repeated over time to identify changes.

What types of content perform best in AI search?

There isn't one universally successful format. Our snapshots highlighted useful specialist pages, practical guides and detailed service information, while wider 2026 research shows citation preferences vary by industry and platform. The priority should be content that answers relevant questions clearly, accurately and with supporting evidence.

Does improving GEO guarantee more recommendations?

No. AI systems generate answers dynamically, and visibility can vary with the question, platform and information available. GEO helps businesses understand and address potential gaps, but neither citations nor rankings guarantee recommendations, enquiries or sales.