The AI discovery shift x Similarweb

Missed the live event? Watch the full recording above, then read on for the key takeaways from The AI Discovery Shift. 

At our recent webinar, The AI Discovery Shift, Balance co-founder Alex Murphy and Natasha Chatterjee, Principal Consultant at SimilarWeb, shared fresh data and practical guidance on how generative AI is reshaping the way consumers find and research financial products. Here’s a breakdown of what you need to know. 

AI is changing how people search for financial products

SimilarWeb’s Natasha Chatterjee opened with data that showed monthly Google volumes for mortgage-related keywords over the past 12 months were measurably lower than the year before. Her and the team at Similarweb, showed this chart below represents this indexed trend: 

Generative AI is solving a genuine consumer need before they try even need to go to a search engine for a comparison site. The average Google search is around three words. The average ChatGPT prompt is nearly 70. People aren’t just asking shorter, sharper questions, potential customers are having full conversations bespoke to their own situations and getting bespoke recommendations.  

These are richer, more personalised answers, all within a chat sessions without needing to visit a range of different comparison sites. And as a result, traditional search volume is stagnating while AI conversation volumes are climbing fast. 

But don’t be mistaken – Google is still very much the dominant channel, that has not changed. But when you index from April 2025 as a baseline, the trajectory is clear: Google is plateauing, and AI tools including ChatGPT, Gemini, Perplexity, and Copilot are growing significantly. 

What this means for financial services brands specifically

This isn’t a general internet trend that happens to touch financial services. But it is hitting the sector directly due to intermediated way people shop for complex products, and the SimilarWeb data made that visible in a way that’s hard to ignore. 

When looking at the UK’s top finance sites and measuring where their referral traffic comes from, ChatGPT has become the second largest source of traffic for the sector. Sitting ahead of price comparison sites and financial advice platforms that have historically dominated referrals.  

Even if we consider that a lot of this search could have mixed commercial intent than PCW traffic (which we’ll see in a moment), that’s a significant shift in a very short space of time. On top of that, the breakdown of AI queries within financial services is also telling.  

Savings and investments account for the largest share of prompts (around a third), followed by trading and currency, then mortgages and lending. Insurance, accounting, and daily banking queries make up the rest.  

Claim management (there’s that low commercial intent audience) alone represents around 22% of insurance-related AI conversations, which raises an important question for insurers: are policyholders going to ChatGPT instead of your website to understand their cover? And if so, what are they being told? Are you actively managing this regulated messaging? 

There’s also a quality dimension to this traffic to bare in mind.  

Conversely, data from the e-commerce category (used as a broader proxy) shows that, aside from paid search, ChatGPT is outperforming organic channels on conversion rate.  

These are the high-intent users arriving further down the funnel, people who’ve already done their research through conversation and are ready to act. That’s represented by the graph below that found that ChatGPT converted twice as well as organic search, social media and email. 

Not all brands are winning equally

Perhaps the most important slide from Natasha’s section was the one showing AI referrals indexed against overall web traffic. When you adjust for the size of each brand, a clear pattern emerges digitally native, app-first, review-led brands are receiving a disproportionately high share of AI referrals. The brands that are actively optimising for visibility are already pulling ahead. 

Similarweb, ChaGPT LLM citations, Feb 2026

Equally important: different LLMs are behaving differently.  

ChatGPT and Perplexity tend to cite brands less frequently than Google’s AI Mode or Gemini. Some brands over-index significantly on Gemini but score low on ChatGPT. This matters because it means there isn’t a single strategy for optimising for AI. Some are going to respond well to search tactics, others more to social tactics. 

You need to understand which platforms your audience is using and how each one is treating your brand. As Natasha put it, don’t think of Gen AI as a monolith. 

Practical steps to improve your AI visibility

Alex, co-founder of Balance, picked up from the data to walk through what it means in practice and what hasn’t changed from modern marketing strategies. 

Good SEO is good GEO

Danny Sullivan, Google’s Director of Search, said it recently, and Alex echoed it throughout: strong SEO puts you in a better position than most when it comes to AI visibility. This was reiterated again mid-May when Google said GEO is basically SEO (though there is still debate). 

Technical setup, well-structured content, URL visibility, topical authority are the fundamentals, and they haven’t been made redundant by AI. If anything, they’ve become more important. The brands showing up consistently in AI overviews tend to have strong E-E-A-T signals, solid backlink profiles, and content built around expert insight. That’s not a coincidence. 

Keyword research is becoming topic ownership, because when someone types 68 words into ChatGPT, you can’t optimise for a head term. You need to own a subject and multiple dimensions of it so you can cover a consumers various category entry point.  

Measurement is also becoming less “clean”; the direct input/output visibility of traditional SEO rankings doesn’t translate neatly to AI referral tracking. This is probability engine, and probabilistic results change constantly. Get comfortable with that ambiguity and start building new frameworks for assessment.  

Digital PR is no longer optional 

Your digital PR strategy needs to be treated as a core channel, not a nice-to-have.  

Earned media dominates AI citations. Brand mentions, even without follow links, contribute to the semantic associations that help LLMs understand what your brand stands for and what it should be recommended for. 

Alex’s advice on the tactical side was specific.  

Fix broken inbound links, they’re often pointing from high-quality sources that no longer send equity anywhere but were in the training data.  

Revisit content that performed well but wasn’t given enough distribution.  

Build evergreen assets that are uniquely yours.  

Think about how you describe your brand in press releases and outreach, because LLMs are building entity associations from that language. Prioritise editorial list coverage, best-of and top-five articles are heavily cited by AI because they match the format AI uses to answer comparison queries. `

The correlation between brands with strong digital PR profiles and brands winning in AI visibility was a recurring theme throughout Alex’s section, and it’s backed by the referral data Natasha presented.  

Don’t use AI to write your content 

This came up twice, once in the main presentation, and again in the Q&A. The question of whether AI-generated content, with a named author added, can still perform is one the team gets asked regularly. 

Alex’s view is clear: it’s a high-risk tactic. There’s a phenomenon he described as “rank and tank”, which is where AI-generated content can produce an initial traffic spike, followed by a sharp drop when algorithms catch up.  

More fundamentally, LLMs are designed not to recycle AI-generated material. They’re looking for fresh data, expert insight, and proprietary thinking,  the things only human expertise can genuinely produce. Use AI as an editing or research tool by all means. But fact check and plagiarism check like crazy and invest in writers who know your subject. 

Don’t panic – but don’t stall either

Alex closed with a mentioning that there’s a lot of noise in the market about AI tactics, many of which are debunked within weeks. The team’s advice is to ignore the FOMO, be selective about your sources, and focus on sustainable fundamentals rather than quick fixes. 

That said, quick wins do exist. Refreshing existing content and earning new coverage can produce AI visibility results within days. You don’t have to overhaul everything at once, but you do have to start. 

Want to go deeper?

The webinar covered a lot of ground in a short time, and there’s more to explore, particularly around social media’s growing role in AI discovery. 

If you’d like to talk through what any of this means for your brand specifically, we’d love to hear from you.