This article is taken from/inspired by the video I recoded for OnCrawl recently as part of their AI myths vs data series.
AI can render content in a browser - it is pretty good at doing it now. It can also read text in an image and can interact with JavaScript-rendered content. Does that mean AI Chatbots or other search surfaces will do this every time? That’s an assumption that’ll create risks.
AI Chatbots can do a lot - but will they?
AI pipelines all have different remits, different processes and tool-calling capabilities. What works for one AI “surface” or interaction point may not work for another. Another of the significant shifts between “traditional” search vs what we’re working with now.
Take images as an example, AI can “read” an image very easily if you upload one and prompt it to read the text. At least that is what I found when I tested it here.
That does not mean an image containing important text will be read and interpreted as part of a search or retrieval pipeline. If that text was needed to make you eligible for an answer, you cannot be confident it will be seen at this point. The capability exists for this to happen, the question is whether this process is invoked. The same applies to other more costly elements of an agent’s activities, like firing up a browser and rendering client-side resources.
Many agents with browser capabilities can render JavaScript content just like a normal user, they typically don’t do it all-that often. In “agentic” scenarios rendering only happens if the system decides it is really necessary. Or when the user explicitly asks it to - which will be very, very rare.
Most agents will do as little work as they can get away with.
So they’re fundamentally lazy, they may fetch the page, they may inspect the returned content and they may look at the accessibility tree. But fully rendering client-side JavaScript and inspecting what appears afterwards is one of the least likely results (for now).
That creates a fairly simple risk, that if important information is buried inside images or depends entirely on client-side rendering, you are relying on an additional processing step taking place before that information can be discovered or used.
Sometimes it will, sometimes it will not. But if that (hidden) information important to the customer journey, discovery or retrieval - you may not be shown.
The safest approach is still the one SEO has been recommending for years: if information is genuinely important, make sure it exists in the raw HTML served by the page.
AI systems may be capable of doing considerably more than traditional search crawlers. But capability is not the same thing as guaranteed execution.
If the information is mission-critical, do not depend on an optional extra step in the pipeline to make it visible.



