Platforms · Software · Providers · Markets
AI assistants have become a meaningful route by which prospective clients discover and assess financial-services brands, says Andy Francos, Growth Consultant at Pepperstone. He says a response to a broker-selection question can present three or four names before a user visits any firm's website. That makes a broker's visibility dependent on more than the pages it owns.
Zay Yang, Founder & CEO of Wordflow, says the information that influences an AI recommendation is often found outside a company's own site. Wordflow helps brands monitor and improve how generative AI platforms present them. For broker operations teams, that raises practical questions about whether licensing and fee details are readable, correct and represented consistently wherever an engine retrieves them.
Francos says Pepperstone is trying to secure mentions across several AI products, while Yang points to third-party citations and the quality of a broker's own web content. The shift affects content management and compliance as well as marketing: an answer that misstates a licence or charge can shape a prospective client's view before the firm has a chance to explain it.
The discovery systems brokers must monitor
- Engines draw different audiences. Francos says Pepperstone monitors ChatGPT for its reach, Google AI Overviews and Google AI Mode for their presence in browsers and Android devices, Gemini for similar reasons, and Claude for business research. The source reports about 1.2 billion weekly active users for ChatGPT and over 50% of the AI search market share. Google AI Mode has more than a billion monthly active users, while Google AI Overviews reach over 2.5 billion monthly users.
- Answers depend on the question. Yang says a query about a broker's authorisation tends to lead AI systems to the relevant regulator, not to an endorsement by that regulator. A question about a firm-specific detail, such as a deposit bonus, may instead send the system to the broker's site. Yang says those details need to be available as ordinary text: JavaScript-heavy displays and PDFs may not make the information accessible to the engines. He also says FAQ and entity markup can help systems identify a firm and reproduce details accurately.
- Outside references can carry more weight. Francos says third-party pages make up a significant portion of citations associated with Pepperstone in Wordflow's broker data. He says those sources are often cited ahead of the firm's own website. The mix varies by location: examples include ASIC and MoneySmart in Australia, the FCA, Trustpilot and BrokerChooser in the UK, and MAS and local finance publications in Singapore. Yang says the citation sources across these markets have little overlap.
- Reddit's influence is not uniform. Reddit made agreements with Google and OpenAI in 2024, but Francos says its position in Pepperstone's broker prompts has changed. After GPT-5.6 reached general users, he says Reddit disappeared from the firm's reference list, even as total citations quadrupled. In UK broker prompts tracked by Obsero, Francos puts Reddit at 32nd, behind the FCA, gov.uk and competitor sites. He says Pepperstone's paid Reddit activity did not produce a return and that community engagement matters more than treating the platform as an advertising outlet.
Why local citations reshape answers
Wordflow's findings, as reported by Finance Magnates, point to a difference between being visible in one country and being understood consistently across several. Yang says English-language training data and references are more prominent than material in other languages. A broker may therefore receive confident treatment in one market and more qualified coverage in another, where the local sources used by AI systems differ.
Yang calls the commercial effect of those qualifications a hedging tax. A recommendation can still be weakened by caveats about a firm being newer or having higher fees. He says local sources may be less contested than those in the US, giving brokers a chance to establish a presence more quickly. Francos also distinguishes between a model's older training knowledge, which he says changes through sustained brand-building, and live retrieval, where public relations, partnerships and current content can affect citations sooner.
Discovery also complicates attribution. Francos describes a journey in which a trader reads a forum discussion, asks ChatGPT for regulated broker options, searches independently and then opens an account.
The broker may record only an organic brand search, leaving earlier influences uncredited. Separately, James Buckley-Thorp, founder and CEO of UK-based AI company Atlian, warns that planting Reddit discussions to prompt AI recommendations would count as financial promotion. He links the concern to the FCA's statement that 80% of customers lose money.
What this means on the desk
Operators should check whether licensing, charges and account details are available as readable text, and whether those details match the information held across regulator and comparison sources. They should also ask which sources appear in AI answers for each market and language they serve. The source gives no operating-cost figures or implementation timetable, so those remain questions for teams planning changes to content systems and monitoring.
There is also a longer-term integration question. The industry has begun testing Model Context Protocol (MCP) connections between trading accounts and AI agents; the source says current safeguards stop agents from placing trades autonomously.
Teams should establish what account data those integrations expose, what controls apply and how any future changes affect execution systems. The article describes a possible shift in which a trading app becomes an execution backend, but does not specify how or when brokers would make that transition.
