A significant shift is occurring in the world of large language models (LLMs). Until recently, these models have relied on using information from historical training datasets. The next evolution, which is already becoming more prevalent, is ‘runtime search’. Rather than replying with frozen, static information – generating responses based on statistical patterns drawn from something that was input possibly months or years ago – AI models are now becoming tools that research in real time, scouring the internet at the very moment a user prompts them. This development reinforces the traditional wisdom about the importance of the first page of search results: if LLMs are leaning heavily on real-time web retrieval, the most prominent information will anchor responses – and it is more important than ever to manage the information about you online.
Historically, at the foundation of how an LLM handles facts is parametric memory: the data taught to the AI model during its training process. The reliance on this is also what causes some of the most common frustrations people experience with AI. One of the technology’s key limitations is rapid data obsolescence – the knowledge of an LLM is frozen in time. In the real world, critical information, such as live financial data, updated medical guidelines, or legal precedents, needs to be current if LLMs are to be effectively deployed. Businesses need accuracy today, not at the time of training.
Fixing the limitations of parametric memory through more frequent retraining cycles is unsustainable, both logistically and fiscally. Retraining is also slow to deploy and difficult to verify. The rate at which tech companies are spending on AI is well documented, but even those with the deepest pockets and the most bullish of investors cannot afford to retrain their models every time the news cycle changes – hence the shift to runtime search.
New AI models are using new architectures to address this. Advanced Retrieval-Augmented Generation (RAG) systems are designed so that, when a query is submitted, the LLM initiates a search query and retrieves relevant external information from the web, databases, document stores, APIs, or search indexes. It then assesses the results, and autonomously refines its own search queries to look for further sources, while filtering out unreliable data. Consider a user asking about an individual; rather than drawing on potentially outdated training data, the model searches live sources and articles, evaluates their credibility, and synthesises a response from what it finds in that moment. Test-Time Compute, meanwhile, allocates more processing power during this inference stage, allowing the model to reason more carefully – checking its own logic, weighing competing claims, and arriving at more considered conclusions. A way to conceptualise this is that we are transitioning from models that remember to models that reason.
This has direct consequences for how reputations and narratives are managed. Because AI will attempt to provide an answer swiftly, it will lean on the most prominent URLs – typically the first page of search results – which effectively become the ingestion channel for LLMs. The caveat to this is that LLMs may search for multiple additional things perhaps not considered by a human researcher and weight these accordingly, but the process is not limitless and there will be a bottleneck. The sources that get relayed will share many characteristics with those that rank well in conventional search engines: they will come from authoritative domains, be presented in a clear and digestible structure, and be frequently cited by other pages. It is worth noting that runtime search models also weight recency and internal consistency – prominence alone does not determine what surfaces – but the primacy of well-ranked, authoritative sources remains the dominant factor.
This creates both opportunities and challenges. Because runtime search draws on the live online landscape, a person or company can influence AI output by shaping what exists on the web in real time. With a clear, optimised communications strategy, a correction or authoritative statement could be reflected in an LLM’s response within hours, or even minutes.
However, the reverse is equally true. Hostile third parties could deliberately poison the online record to produce a negative LLM output. But it does not even require an active bad actor – a well-reported crisis, a regulatory issue, or unbalanced media coverage can be ingested immediately and presented as a highly persuasive summary, even to users who prompted with entirely neutral language. Traditionally, a user might read a range of articles, researching different sides of a topic, but AI could diminish this greatly as it will cede undue influence to a user’s favourite AI-powered search chatbot, which compresses multiple pages into a single answer and reduces exposure to source diversity. The speed and authority with which AI delivers these syntheses makes the underlying source environment more consequential than ever before.
To combat this, reputation management needs to be proactive – it cannot just wait for a crisis to blow over. Taking care of one’s digital hygiene is vital. What this looks like in practice is ensuring the live web landscape features high-authority, optimised, structured data. In the event of a crisis, the Advanced RAG system needs to be able to ingest corrections alongside the crisis narrative. In turn, this allows the Test-Time Compute mechanism to weigh both sides fairly. This could include publishing press releases on a credible domain, correcting Wikipedia entries, or maintaining well-labelled metadata content; all of this can feed directly into what an AI model retrieves and presents.
The emergence of runtime search is positioned to potentially amplify the traditional best-practice rules of reputation management and digital communication. The shift will reward those who have built something accurate, up to date, and clearly structured. Authoritative content, credible sources, and a well-maintained digital presence are going to remain indispensable. What will change is the consequence of getting it wrong. When a human user reads a damaging article, they might seek a second opinion; when an LLM ingests one, it summarises it into a confident, neutral-sounding summary with no asterisks, no invitation to question the source, and no hesitancy. Individuals and companies need to treat their online presence as a live information environment to be actively maintained. In a world where AI becomes the first port of call, that environment is their reputation.
Digitalis
We firmly believe that the internet should be available and accessible to anyone, and are committed to providing a website that is accessible to the widest possible audience, regardless of circumstance and ability.
To fulfill this, we aim to adhere as strictly as possible to the World Wide Web Consortium’s (W3C) Web Content Accessibility Guidelines 2.1 (WCAG 2.1) at the AA level. These guidelines explain how to make web content accessible to people with a wide array of disabilities. Complying with those guidelines helps us ensure that the website is accessible to all people: blind people, people with motor impairments, visual impairment, cognitive disabilities, and more.
This website utilizes various technologies that are meant to make it as accessible as possible at all times. We utilize an accessibility interface that allows persons with specific disabilities to adjust the website’s UI (user interface) and design it to their personal needs.
Additionally, the website utilizes an AI-based application that runs in the background and optimizes its accessibility level constantly. This application remediates the website’s HTML, adapts Its functionality and behavior for screen-readers used by the blind users, and for keyboard functions used by individuals with motor impairments.
If you’ve found a malfunction or have ideas for improvement, we’ll be happy to hear from you. You can reach out to the website’s operators by using the following email webrequests@digitalis.com
Our website implements the ARIA attributes (Accessible Rich Internet Applications) technique, alongside various different behavioral changes, to ensure blind users visiting with screen-readers are able to read, comprehend, and enjoy the website’s functions. As soon as a user with a screen-reader enters your site, they immediately receive a prompt to enter the Screen-Reader Profile so they can browse and operate your site effectively. Here’s how our website covers some of the most important screen-reader requirements, alongside console screenshots of code examples:
Screen-reader optimization: we run a background process that learns the website’s components from top to bottom, to ensure ongoing compliance even when updating the website. In this process, we provide screen-readers with meaningful data using the ARIA set of attributes. For example, we provide accurate form labels; descriptions for actionable icons (social media icons, search icons, cart icons, etc.); validation guidance for form inputs; element roles such as buttons, menus, modal dialogues (popups), and others. Additionally, the background process scans all of the website’s images and provides an accurate and meaningful image-object-recognition-based description as an ALT (alternate text) tag for images that are not described. It will also extract texts that are embedded within the image, using an OCR (optical character recognition) technology. To turn on screen-reader adjustments at any time, users need only to press the Alt+1 keyboard combination. Screen-reader users also get automatic announcements to turn the Screen-reader mode on as soon as they enter the website.
These adjustments are compatible with all popular screen readers, including JAWS and NVDA.
Keyboard navigation optimization: The background process also adjusts the website’s HTML, and adds various behaviors using JavaScript code to make the website operable by the keyboard. This includes the ability to navigate the website using the Tab and Shift+Tab keys, operate dropdowns with the arrow keys, close them with Esc, trigger buttons and links using the Enter key, navigate between radio and checkbox elements using the arrow keys, and fill them in with the Spacebar or Enter key.Additionally, keyboard users will find quick-navigation and content-skip menus, available at any time by clicking Alt+1, or as the first elements of the site while navigating with the keyboard. The background process also handles triggered popups by moving the keyboard focus towards them as soon as they appear, and not allow the focus drift outside of it.
Users can also use shortcuts such as “M” (menus), “H” (headings), “F” (forms), “B” (buttons), and “G” (graphics) to jump to specific elements.
We aim to support the widest array of browsers and assistive technologies as possible, so our users can choose the best fitting tools for them, with as few limitations as possible. Therefore, we have worked very hard to be able to support all major systems that comprise over 95% of the user market share including Google Chrome, Mozilla Firefox, Apple Safari, Opera and Microsoft Edge, JAWS and NVDA (screen readers), both for Windows and for MAC users.
Despite our very best efforts to allow anybody to adjust the website to their needs, there may still be pages or sections that are not fully accessible, are in the process of becoming accessible, or are lacking an adequate technological solution to make them accessible. Still, we are continually improving our accessibility, adding, updating and improving its options and features, and developing and adopting new technologies. All this is meant to reach the optimal level of accessibility, following technological advancements. For any assistance, please reach out to webrequests@digitalis.com