Serve Markdown to AI Agents with Accept Headers
acceptmarkdown.com
acceptmarkdown.com
> It is a bad design trade-off to send a bunch of header fields on every request just to tell the server all of the possible variations of preference held by the user, particularly when there is a very small chance that any of those dimensions are applicable to the target resource. It has been a bad design trade-off ever since the very brief period in 1993-94 when folks didn't know which image format would be usable on all UAs and there was no CSS or javascript to allow for client-side adaptation.
> ...The caching impact of proactive negotiation is far worse than the one extra round trip per site for reactive negotiation, and even that round-trip isn't necessary in formats that support client-side adaptation.
On the caching impact, Simon Willison wrote:
> ...you can’t deploy an application that uses content negotiation via the Accept header behind the Cloudflare CDN — for example serving JSON or HTML for the same URL depending on the incoming Accept header. If you do, Cloudflare may serve cached JSON to an HTML client or vice-versa.
Note: I posted this in another comment with links to those two quotes which I couldn't copy easily now - will add later.
I take the point that it makes caching harder, but I don’t think that should overrule ergonomics concerns.
But for HTML and Markdown in particular, there's been so much useful work done in the semantic HTML space and microformats, that I don't know why anyone interested in this wouldn't just improve their HTML markup and leave it to the agent to do the rest. Convert to markdown or extract the useful HTML before handing it to model.
It's different content representations. A text in a markdown file is conceptually the same content as the same text in HTML (or PDF).
Link: /some-page rel="canonical"
Link: /some-page.json rel="alternate" type="application/json"
Link: /some-page.html rel="alternate" type="text/html"
Link: /some-page.md rel="alternate" type="text/markdown"On the flip side, I'd argue that the current centralisation of user agents (in the classical sense here) that benefit from programmatic content negotiation in form of a handful of harnesses like Claude or Codex is a great lever toward forcing the ecosystem to adopt better practices: If Anthropic added content negotiation as described in this thread to Claude, many sites would be incentivised to improve their web servers.
Maybe it will notice those, and maybe it will figure out the pattern for follow-up page requests, but there's no guarantee and it won't help the first request.
I am well aware that few sites are taking that much care of their API in terms of HTTP features, but all of the problems discussed here have solid and battle-tested answers.
The content at a URL should always match, the format in which its represented can be different based on the request. Its a bit like buying a book in hard copy or paperback, same book different format.
Content negotiation still has its uses, but most of the time you’re better off using different endpoints.
Akamai is another case where Vary is harmful, from their docs [1]:
> As the content in response may be different for the same URL,
> Akamai edge servers don't cache responses that include the Vary header,
> even if the content is cacheable by definition. The only exception is
> the case where the Vary header's value is Accept-Encoding and the
> Content-Encoding header's value is br or gzip – edge servers cache
> such responses, applying the caching rules you defined in your property.
https://techdocs.akamai.com/property-mgr/docs/rm-vary-header
Wrong: https://developers.cloudflare.com/cache/concepts/vary/
Simon wrote that in 2023: https://simonwillison.net/2023/Nov/20/cloudflare-does-not-co...
But it's something I think developers should think about if they rely on caching. If it took Cloudflare this long to support this, there may be other systems which still don't.
Link to Roy Fielding comment:
https://lists.w3.org/Archives/Public/ietf-http-wg/2013JanMar...
Seriously?
Doing this with the Accept header is a bad idea, although I think CSS and JavaScripts (in web pages) is not a good solution to this either (they can often make it worse).
My way is the Scorpion conversion file, which must be downloaded explicitly by the end user and the end user must be allowed to override it with their own, and which tells it what to do when it receives a file that it does not recognize, based on the URL or the file type, such as: rewrite the URL, use a uxn program to convert it (to a format that you can use), use a uxn program to display it, etc. Something similar might be possible to add into WWW, by adding a "Interpreter:" response header into HTTP, perhaps using WebAssembly instead of uxn.
It was already optimized for bots and search engines (which are bots) and it has been used for decades. Why we need to serve in markdown now?
There are also many parts of the HTML, like navs, that are useful for bots and AI and may be removed in the markdown version.
That's effectively using an LLM as an HTML to markdown converter, which is both absurdly wasteful and also surprisingly inexpensive (if you use a model like GPT-5.6 Luna.)
This is kind of the opposite of reality no? Markdown is just plain text and meant to be human readable. You don't need XML tags to read and write it, opposed to html where you do and you need a browser to properly view it.
HTML I can view in any browser / webview / etc. Good markdown viewers are fewer / more special, or end up translating md to html for display.
And by composing, I didn't mean writing by hand. We are talking about prompting, right? Or that is what I thought we are talking about. Composing HTML "components" into a final prompt HTML is easier than composing markdown snippets into the final prompt. That is because with HTML there are several ergonomic libraries to parse HTML to AST and to format AST back to HTML. The libraries (for parsing to AST and back to strings) are more limited with markdown.
Of course, if you build up your prompt entirely by hand (copy pasting snippets, etc.) you can get away with markdown or plain text. That works okay for some harnesses, but it leaves too much control (or more accurately, opportunity to misunderstand) to the harness.
Assuming that is what we are talking about, HTML is easier to work with than Markdown, unless you are writing it by hand. That is, composing semantic HTML is more ergonomic than composing a Markdown formatted document from components / snippets, programmatically. The libraries are just better and more versatile in most programming languages. Typically you go from HTML or Markdown to AST, then you compose them to end up with the final tree, then you format the tree to HTML or Markdown. LLMs treat them basically identically (context in HTML or context in Markdown), so I have ended up forming complex prompts / context parts using HTML.
I'm not seeing this at all. I've got a small search engine I made that strips HTML back to Markdown for its full-text indexing. HTML is typically 10x bigger than the Markdown of the actual content, but that's because the majority of HTML out there is truly terrible.
I personally like HTML, and my own webpages are all hand-coded HTML. In that case, it's probably a closer ratio to what you describe. I'd suggest it's much higher than 20% more, but it's not likely more than double. But that's assuming someone paying attention to the efficiency of the HTML, and most people / websites just don't.
Markdown is even more readable without tools than HTML - it's essentially a plain text document - but I agree that HTML is better for actual semantic structure.
I am comparing expressing information in semantic HTML or in Markdown. HTML doesn't bloat it much, and it is easier to work with (programmatically in python, ts, go, java, etc.; including parsing and composing and formatting). I am not talking about web apps, which commonly have more "app" than "information".
And the ratio is really closer to 1.1x tokens than 1.2x, unless the information you are expressing is highly structured. Typically 1 structural element (html tag, markdown marks such as ###) ends up being each 1 token. HTML often has the matching closing tag, and that is really what the difference in number of tokens boils down to. Feel free to run your own tests and compare (I ran a few and it was within 1.09-1.14x).
https://www.abc.net.au/news/2026-08-27/shania-twain-intervie...
This HTML page is 360523 bytes of raw HTML. After it has gone through my Markdown parser, it is just 9915 bytes of plain text / Markdown, about a 97% reduction. Only 3% of the HTML is actual content.
The point of Accept Markdown is to save web hosts bandwidth. An AI harness can (and already does) download the HTML & parses out a Markdown version so that it is only minimal tokens before it hits the context window. But I still need to download the 360KB of HTML from the server in order to extract the 9KB I actually need. By serving Markdown versions of your page to AI agents, you can save 97% of the bandwidth that AI agents might be incurring.
There's no reason this page needs to be 360KB of HTML. A handcoded / handoptimized HTML file might be only 12KB - converting from Markdown to HTML is only going to minimal file increase. But that isn't what the web is. It's full of slop generated by CMS applications & Bootstrap & web frameworks and relics of Frontpage edited WYSIWYG HTML editors. Accept Markdown is trying to get webhosts to save everyone bandwidth by serving the Markdown from their side. Maybe it has a chance if it gets baked in at the web server level, or because Cloudflare is applying it to sites that flow through their network. But I think it's ultimately futile - the same people who don't know their Wordpress output is garbage, also won't know how to configure Markdown on their server.
When I am providing information for humans, I CAN go with semantic HTML (that 9kB Markdown becoming maybe 10KB semantic HTML). I can also go with "framework" layers and turn that into a massive 1MB crap which when rendered by the end user's browser maybe shows most of my content plus ads plus supply chain attacks I unintentionally pulled through with one of the "framework" layers, ahem, nodeJS, ahem.
When I am providing data for both humans and AI agents, XML and other formal data formats might be good choices. If the data happens to be mostly text, then HTML and Markdown both work, but HTML is easier to compose (with programs). If you don't need composition at all, markdown or plain text are obviously easier to hand-craft.
Only in the last scenario, where you don't need to combine information and are directly writing snippets of text, Markdown comes out as a decent option. I honestly don't have a clue how often that happens in practice. Maybe for casual writers (which may be a big market that I am blind to)? As soon as I want to put in a plain table in the data, Markdown gets in the way.
Which React package is this?
There's no "agent". It's a few wrappers around API calls in a trenchcoat.
Better idea: make your UI always prioritize the content.
It's more along the lines of "Oh, hey there, you look like a bot, here are details on how to use our monetized mcp".
More likely though is that nothing will change here, because people will abandon it as soon as there is a paywall.
I also think it's exceedingly unlikely that any of the top 4 chatbots would choose to load websites this way. Too many risks for a 0.01% adoption even years later.
HTML is a markup language. User agents present it in a way that makes sense for the user—visually, or through assistive technology. Nothing about adding LLM “users” should change this—their user agent, the harness, should present HTML in a way they can natively understand, by translating it to Markdown.
We are not going to rewrite the entire web because harness developers are too lazy or stupid to pull an HTML to markdown package from npm. If some sites want to do so anyway, good for them, in many cases I’d love to skip the CSS/JS and read the markdown directly (or better, nicely formatted). But don’t blame website authors for your harness wasting your tokens.
I've been enjoying https://defuddle.md since learning about it. Works great. Not affiliated.
Most harnesses have a "zero config" plugin that works, eg: https://pi.dev/packages/pi-exa
(no affiliation to either)
The reason it may be more difficult today is because we've lost a lot of that. Some of it because of modern JS frameworks, but some also because publishers simply don't want to make it easy for the useful stuff to be scraped and extracted easily.
I'm not convinced that's changing because of AI agents (it's getting worse in many ways with anti-agent rules). Maybe improving for documentation pages intended for agents. But if it is changing, I think it'd be far easier to improve the HTML and let the agent take care of the rest.
I'd even say, is that even a problem ? If your website isn't a front end blob soup and just a relatively clean HTML file, your LLM is totally capable to handle it anyway.
And I'm still wondering myself which incentive do I have to serve my content to LLMs anyway.
If you're working with a marketing team, you'll find that they are desperate to serve their content to LLMs, to counteract the rapidly shrinking number of people actually visiting websites. What I don't want is for that to mean "build the website twice, one for humans and one for LLMs". Websites are for people—bots can adapt.
> Why do you need AI then?
(from this HN comment, worth a read in full: https://news.ycombinator.com/item?id=43679585)
It's giving https://news.ycombinator.com/item?id=8863#9224
Except it isn't. Someone else could build that stupid frontend, and essentially create a better booking.com. Or you could ask the robot to make you a stupid frontend. It remains true that a well-designed MCP server goes wildly against the interests of sites like booking.com.
Even something as simple as a recipe blog. That one is already solved by chatbots, and it’s glorious. I never have to look at a shitty recipe site full of ads with the recipe buried beneath a mile of fake stories again. LLM finds some, reads reviews, I pick one, if it’s good import to paprika, and that’s that.
There’s at least a chance that enough people demand it, combined with AI hype, that we could end up with a slightly better experience.
And if not, then there’s always the brute force approach of something like grok bot
A cherry.jpg image/jpeg file is shown as an image by the browser
A logo in svg like https://news.ycombinator.com/y18.svg too
So browsers should show readme.md as text/markdown and clicking on a link inside that file should also redirect to another markdown file, instant wiki
Browsers should implement that asap and we will follow
Practically, Markdown is more of a vibe than a spec and everyone just uses whatever subset makes sense to them.
They’re far too useful for me.
The open internet deserves a real solution instead of gatekeepers in the form of brightdata, firecrawl, cloudflare and the likes.
OTOH, serving ads to mistrain AI content is an interesting business model.
What you _should_ be doing is ensuring that your page works with screen readers. That simultaneously solves any kind of machine-readability issues and also makes your website accessible for the blind.
The MD content isn’t intended for human consumption whereas HTML is. So you either have that injection readable to your users, which could be jarring to non-technical readers, or you play the cat-and-mouse game of hoping those pesky crawlers don’t ignore text that’s not human-visible on a rendered page.
Or everything just disappears behind something like cloudflare or a paywall.
The problem is that it’s in the sites interest to be SEOed. But it’s against their interests to be AI crawled.
Maybe tarpit them into reading gigabytes of Markov Chain nonsense, really slowly?