AI search in the capital markets
⏲ Read time: 5 minutes
While marketing departments are adapting the commercial website for AI search, the IR function still relies largely on static PDF archives. But as investors and analysts now hand over their initial research to AI agents, a financial risk emerges: if your primary data cannot be read directly, secondary sources get the final word.
Ask your CMO what is on the agenda right now, and AI search optimisation, above all through the GEO (Generative Engine Optimisation) methodology, is very likely to be high on the list. The way information is searched for and consumed has changed fundamentally: traditional search is rapidly being replaced by AI models such as ChatGPT, Claude and Google AI Overviews.
This changes the playing field in two decisive ways.
First, AI crawlers are not looking for isolated keywords and phrases, but for structured knowledge: so-called entities, context and verifiable facts.
Second, contact with your information now takes place through an intermediary. Users explore, ask questions and evaluate facts about the company directly in the AI model, before visiting the company’s website, and sometimes without ever doing so. The AI agent compiles the information, draws its conclusions and shapes the reader’s first impression. If the company’s data is not available in a format that algorithms can easily read, interpret and cite, it risks being excluded from the answer or described on the basis of secondary sources. It is therefore no surprise that marketing departments are investing heavily in evaluating and adapting their digital presence.
But this is no longer solely a question for the commercial business. The same shift is now taking place in the capital markets, with far-reaching consequences for digital IR.
Most analysts and investors already use AI in their research process, and trust in the technology is strikingly high. In Brunswick’s 2026 U.S. Investor Survey, 44 percent said they trust AI-generated summaries of financial material as much as reports from traditional sell-side analysts. When they ask their AI models to summarise the company’s investment case, analyse margin development or scrutinise sustainability targets, the company’s communication faces an entirely new filter.
In practice, the chain of recipients has been turned upside down: the first reader of your reports is rarely an analyst in the City or a fund manager in Stockholm, but an algorithm that analyses the material in a few milliseconds. Before an analyst has even opened their spreadsheet, the machine has already absorbed the report, interpreted the guidance and passed judgement on the company’s tone. One consequence is that more and more companies run sentiment analyses of their own communication in advance. Ahead of quarterly presentations, scripts and press releases are run through algorithms to identify wording that risks being interpreted negatively by AI models and trading bots before a human analyst has had a chance to put the information in context. Linguistic nuances, which were previously calibrated to land well with human readers, have become a semantic game of chess against an algorithm.
This development also means that companies’ traditional reporting formats face a review that is both necessary and urgent. For understandable reasons, financial reporting has historically focused primarily on compliance, accuracy and regulatory certainty. Reporting processes have been designed around proven, statutory formats such as PDF documents and formal ESEF packages. It is a logic built for control. But in a digital environment where algorithms are the first readers, this way of working creates unforeseen friction, as the traditional formats make the information hard for AI models to access.
When an AI crawler encounters a static PDF, technical obstacles arise. Tables risk being misinterpreted, the logical structure can break down and figures can lose their context. The consequences for the company can be serious: if your primary data is not immediately machine-readable, the AI model is forced to fill the gaps with information from secondary sources. The picture of the company that the algorithm then delivers to analysts and fund managers is, at best, unclear and, at worst, outright misleading.
Step one: Establish AI-readable reporting in open HTML
The first step in overcoming this friction, and establishing genuinely machine-readable reporting, is to give financial communication the same digital conditions as the commercial website. As we discussed earlier in The new paradigm of reporting, financial reports need to be published in open HTML, the native language of AI models. However well designed a PDF report is on screen, it requires extra processing steps for an algorithm. When interim reports and annual reports are published in open, structured HTML, the information becomes immediately searchable, readable and easy for AI agents to cite.
Step two: Maximise the impact through a unified reporting framework
Once the foundation of open HTML is in place, the natural next step is to connect the financial reports with the IR website and the rest of the company’s digital presence. This creates entirely new opportunities to make material information accessible and to show how different aspects of the company’s operations relate to one another. It is through this integration that you get the maximum return on your communication efforts.
When an AI model analyses a company, it looks for confirmation that what is claimed at an overall level is also supported by the figures and notes. If your IR website describes an expansion into new markets, but the interim report’s segment reporting and financial notes sit isolated in an archived PDF, the logical chain breaks for the algorithm. This increases the risk that it misses the context or draws incomplete conclusions.
By connecting IR pages, press releases, interim reports and the annual report in a common digital structure, you create an unambiguous and verifiable source code for the entire company. At the same time, the value for human readers increases. The same architecture that makes your data as accessible as possible to an AI agent also makes it considerably easier to navigate for an analyst, a journalist or an individual shareholder.
Tracking the company’s AI visibility, meaning the extent to which its official communication is cited compared with secondary sources, has quickly become a strategic KPI for the IR function. To reach both analysts and algorithms, IR professionals have started to work more like content producers, applying principles from AEO (Answer Engine Optimisation): the interim report is complemented with modular elements such as short earnings summaries, video clips, searchable transcripts, structured FAQ sections and interactive key figure views. The aim is to provide the machine with verifiable facts before it has time to hallucinate its own picture of the company. This makes data quality the company’s ultimate line of defence.
The financial report has long been regarded as a regulatory end product: a document that is produced, reviewed and filed away. In an AI-driven analytical environment, that view needs to broaden. The financial reports and the sustainability report contain the company’s most thoroughly scrutinised and weightiest facts. Making that information available in an open, linkable format woven together with the rest of the company’s communication is one of the most effective things a company can do to give its capital markets communication maximum impact as the rules of the game are being rewritten.
But a reporting framework in which the corporate website, the IR website, the interim reports and the annual and sustainability report are linked through their natural points of contact is ultimately about more than technology. It requires two things: a robust, well-developed equity story at its core, and an effective division of labour between the different channels. When each one has a clearly defined purpose within the whole, you avoid both unnecessary duplication and fragmented messaging, while regaining control of the company’s financial narrative, whether it is examined directly by an analyst or synthesised by AI.
