WebMCP has the potential to fundamentally change how websites are structured and how their content is consumed.
As I dig deeper, seeing how Agents can allow users to interact with a website’s content, I start to wonder why we would even bother making webpages at all? Why not just have our content available to the Agents, then let the user ask whatever they want?
For decades, we have guessed and re-guessed, studied and re-studied, tested and re-tested, even focus-grouped and re-focused-grouped, trying to figure out how our content should be structured, grouped and separated, into pages, and menus with submenus.
But what hubris, to think we could ever truly guess what users want! And honestly, what ignorance/futility, to think any ONE thing would be what EVERY single user would wants!
When now we are looking at the ability for users to simply ask an Agent for what they want, and get the answer fed back to them in easily digestible bits, no fuss, no muss…
Not some list of search results, links to pages that might have some of the words the user submitted, in some order or another; instead, just like the AI blob at the top of all Search Engine results now-days, the site’s WebMCP Agent would parse the site’s content and return a human-language answer, easily readable by the user.
Example: The user wants to know the store hours on Fridays.
- Currently, a user would type something like “open hours” into a Search bar, hoping that the site editors used a similar phrase on whatever page includes their hours of operation, would get a list of any pages that have the words “open” and/or “hours”, then would need to sift through those pages looking for that content.
Currently, users submit a search query and get a list of possibly useful pages, but not always; and the resulting pages might still require you to bounce between multiple pages to get your complete answer… - With WebMCP, they could type something like “What time do you open on Friday?”, and the Agent would be able to return a phrase similar to “We open at 10am on Friday.”, maybe even adding “Do you need directions to the location nearest you?”
With WebMCP, users might not need to click and scroll through websites to find answers;
they could simply ask a question, and get an answer…
And suddenly, pages and pages and pages of content starts to feel utterly useless…
How many times you have searched for something (maybe before AI search results) and had to repeatedly “click a search result, scan/scroll, click the Back button” until you found, and probably cobbled-together, the data you were searching for? And maybe you’re really smart and hit Cmd+F on pages to speed that scan/scroll part of the process, but you still often had to click-to and click-back a lot to find the answer you were looking for.
Now think how much time/money is involved in designing, developing, QAing, monitoring, troubleshooting, etc., for multi-page designs (Home, PLP, PDP, Blog, Contact, Pages, etc.), across all the various device types and sizes, when perhaps all you could need is a single page with your logo, a text input and an optional Submit button (hitting Enter could also submit your prompt)…
Of course, you would still need all of your content for the Agent to scan, but that does not have to look pretty, just be LLM-readable.
And you would still need a design for the “chat” interface, but that’s basically having a left- and right-justified “bubble”, right?
And maybe this conversation feeds into the bigger topic of why do we even have websites at all, if the search engines are just going to summarize content from across multiple sites into cohesive answers, but for now… let’s just focus on WebMCP, shall we? :-)
Thoughts?
Happy “reading”,
Atg