AI is spurring a seismic shift in the production and delivery of digital content. Websites are losing their importance, and as a result, content management systems are losing relevance too. This change is so radical and profound that it is going unnoticed.
Here’s what has changed: machines are now the most important audience for text. Humans are delegating to bots the work of finding and evaluating details contained in text. And bots need different things from the text than people do.
Humans and bots have different motivations
Most humans are more concerned with expressing what they want than with how an agent should perform a procedure to get them what they want. Humans avoid details if possible, preferring tacit knowledge and heuristic judgment. No one wants to read long terms and conditions statements.
Machines, by contrast, depend on details to avoid messing up. They need lots of instructions and often must synthesize details scattered in many places.
People primarily want to read content that discusses what they want and get — that addresses their intent. They are less interested in the details of how machines interpret those intentions. They will only want to look at the verbose instructions and contracts that machines rely upon if and when there’s a problem.
Humans need articles; bots need files
It’s time to let go of the widespread belief that people and machines need the same things from content. This belief depends on the erroneous idea that bots are just like people and evaluate content the same way. It’s become a popular dogma because it serves as a comforting security blanket as organizations work through the disruptive AI transition.
But it’s absurd to believe that human readability and machine readability are identical. Most web pages are too ambiguous for bot agents to rely on. Organizations need to accept that they’ll have to develop different artifacts for human and machine audiences.
People want human-first articles and messages that are short, easy to read, and may even convey emotion and personality to reflect user intent. Even if bots can fake this style, it doesn’t mean bot agents themselves find such prose useful to act on.
People can be open-ended with their intents; bots perform more reliably when they have concrete direction. People choose AI tools in order to read less.
Agents need machine-first text files: elaborate prose that is full of detail and uses unambiguous terminology, even if the words are technical or legalistic. Bots don’t mind reading more, if it delivers what’s needed. But bots aren’t good at mind reading.
Readers have outsourced the reading to bots
In situations involving information or transactional messages, humans are no longer interested in reading web pages. They’d rather a bot tell them what they need to know; they can follow up if they need to know more. All that text designed to entice readers to notice, click, and take action is being ignored now. Page views and dwell times are meaningless. Web pages no longer have a human audience.
These web pages are increasingly not having a bot audience either. The web page doesn’t tell the bot all it needs to know to make the best decision on behalf of the user. Bots aren’t persuadable, but they need detailed instructions on where to find resources and how to prioritize them.
File management is the new CMS
Many vendors promote the idea that their CMS is now some kind of operating system for bots, as though those human-focused articles are what bots need. In reality, bots need a different set of resources: an array of files, each with a distinct role. The CMS is no longer the gatekeeper.
The CMS will continue to exist to publish web pages that humans want to read. Not all content will be bot-delivered. People will still want to read human-authored content and will visit web pages. But that will represent an increasingly smaller set of the resources consumed on the internet.
Writing for bots as the primary audience
Human writers will be divided into those who write for bots and those who tell bots how to write for humans.
Most human writers will focus on writing instructions for bots. These instructions will be detailed and often collectively written, since bots must rely on multiple files. Apart from the writers developing these instructions, no humans will ever read them. They are bot-centric text files.
A smaller group of human writers will write instructions for bots on how to translate responses into human-agreeable prose, and how to interpret human-submitted prompts (how to interpret teenage slang, for example).
Legacy practices of writing for the website first will change. Humans are no longer the main audience in the AI era. For writers, it will require a shift in expectations, accepting that the text they write in a file may never be read by another person. New metrics will emerge to provide feedback on the effectiveness of bot-focused text files.
— Michael Andrews