[{"data":1,"prerenderedAt":418},["ShallowReactive",2],{"post-\u002Fllms-txt-was-never-the-point":3},{"page":4,"translation":319,"nav":321,"related":404,"random":412},{"id":5,"title":6,"body":7,"categories":295,"category":298,"date":299,"description":300,"draft":301,"extension":302,"image":303,"kind":304,"lang":305,"meta":306,"navigation":307,"path":308,"publishedAt":298,"readingTime":309,"seo":310,"slug":311,"stem":311,"tags":312,"translationKey":311,"type":317,"updated":298,"__hash__":318},"posts\u002Fllms-txt-was-never-the-point.md","llms.txt Was Never the Point",{"type":8,"value":9,"toc":285},"minimark",[10,48,51,54,57,62,65,71,76,79,82,85,96,103,107,110,116,119,122,125,128,132,135,141,156,159,173,176,182,186,189,192,195,198,201,216,222,226,232,238,241,246,249,254,258],[11,12,13,21],"blockquote",{},[14,15,16,17],"p",{},"💡 ",[18,19,20],"strong",{},"Quick Summary (TL;DR):",[22,23,24,36,42],"ul",{},[25,26,27,30,31,35],"li",{},[18,28,29],{},"The point:"," ",[32,33,34],"code",{},"llms.txt"," may be a technical idea worth testing, but it is not an accepted provider standard or a proven AI Visibility lever.",[25,37,38,41],{},[18,39,40],{},"My experiment:"," I tried a more complete information architecture on a suitable site; it was instructive, but it did not create an explosion in visibility.",[25,43,44,47],{},[18,45,46],{},"The bar:"," A hypothesis should not become “add this immediately” before we can explain its conditions, costs, and relationship to provider documentation.",[14,49,50],{},"When a new technical concept is vague enough, there are two ways to approach it. The first is: “That’s interesting. Let’s find out where it helps and where it doesn’t.” The second is: “If you haven’t done this yet, you’re already behind. Do it now.”",[14,52,53],{},"What troubles me about the AI Visibility conversation is how quickly a new idea turns into that second sentence. Before it is even clear what a provider does in which product—before there is an accepted standard, an explainable mechanism, or a measured result—the idea is turned first into a firm prescription and then into a service that can be sold.",[14,55,56],{},"After a while, the attention shifts and the same formula attaches itself to another acronym. There is no need to speculate about what the people behind it know. A technical recommendation does not earn its value from the confidence of the person making it. It earns it when we can explain which product it applies to, under which conditions, and on what evidence.",[14,58,59,61],{},[32,60,34],{}," is a good example of this cycle. When people first began discussing it, my objection was never to the existence of the file. A new format, a new discovery path, a new way to represent content—none of these is a bad idea in itself. My objection was to a proposal with no accepted standard and no provider-confirmed visibility mechanism becoming, in a very short time, “you need to add this to your site immediately.”",[14,63,64],{},"That is why I approach every AI Visibility hype cycle, every tool, and every offer that promises to measure or improve visibility with the same skepticism.",[14,66,67,68,70],{},"This is not a theoretical objection. I actually tried ",[32,69,34],{}," on a suitable site.",[72,73,75],"h2",{"id":74},"i-did-more-than-add-one-file","I did more than add one file",[14,77,78],{},"I did not simply put a single file at the root and point it to a few pages. I built a hierarchical information architecture for several knowledge hubs. I generated Markdown and JSON versions of content pages, then linked to those resources from the root file in an order that made sense to someone trying to follow the subject.",[14,80,81],{},"It was an extra layer alongside the site’s structured data work. Technically, it was enjoyable. Thinking about how a content system presents the same knowledge to a human reader, a web crawler, and different tools is exactly the kind of problem I like working on.",[14,83,84],{},"But I was not telling myself, “AI Visibility is about to take off.” My questions were narrower:",[22,86,87,90,93],{},[25,88,89],{},"Are these resources actually being requested?",[25,91,92],{},"If they are, does the information architecture help a system reach the right part of the content?",[25,94,95],{},"Do alternative representations such as Markdown or JSON have any observable practical effect?",[14,97,98,99,102],{},"Each of those questions is a separate experiment. You cannot reduce all of them to one ",[32,100,101],{},"200 OK"," response, one movement in a dashboard, or one score produced by a tool.",[72,104,106],{"id":105},"the-result-was-not-a-sales-promise","The result was not a sales promise",[14,108,109],{},"The site did not get worse after the implementation. But I did not see an explosion in AI visibility that I could honestly say happened because of it either.",[14,111,112,113,115],{},"That does not prove ",[32,114,34],{}," can never have an effect on any site, under any condition. Drawing a universal mechanism from a single site would simply be the reverse of the mistake I am criticising.",[14,117,118],{},"It does show something else clearly enough: a richer file and content architecture is not, by itself, a proven AI Visibility lever.",[14,120,121],{},"The experiment was still successful for me. Success is not only a line on a chart going up. Learning to ask better questions about a system, seeing the maintenance cost of an architecture, and understanding when I would make the same investment again are results too.",[14,123,124],{},"If I were making the decision today, I would put it this way: if a content system already produces Markdown or JSON representations at low additional cost, this kind of layer can be worth trying. I do not see enough evidence to build a separate information architecture behind an existing system purely because “AI wants it now.”",[14,126,127],{},"This is not an argument against technical curiosity. It is an argument against selling an engineering experiment as a proven marketing return.",[72,129,131],{"id":130},"why-documentation-is-the-starting-point","Why documentation is the starting point",[14,133,134],{},"We do not know every detail of Google’s algorithms. Google does not publish them, and no one expects it to publish its source code.",[14,136,137,138,140],{},"But if Google tells site owners plainly, “We do not use ",[32,139,34],{}," for Google Search; it provides neither benefit nor harm,” then anyone presenting the opposite as certain knowledge needs stronger evidence. The bar for the argument is that simple.",[14,142,143,144,146,147],{},"Google’s current guide to generative AI features in Search is explicit: Google Search does not use ",[32,145,34],{}," or special AI markup, and says the file neither helps nor hurts Google Search. The same guide does not treat prescriptions such as a special “chunking” length or a separate AI schema as necessary either. ",[148,149,155],"a",{"href":150,"rel":151,"target":154},"https:\u002F\u002Fdevelopers.google.com\u002Fsearch\u002Fdocs\u002Ffundamentals\u002Fai-optimization-guide",[152,153],"nofollow","noopener","_blank","Google’s guide to generative AI features in Search",[14,157,158],{},"That does not mean you cannot use the file as an experiment or an organisational layer. It only means that, if you want to claim a specific advantage for Google Search, you need better evidence than what Google itself has published.",[14,160,161,162,167,168],{},"The documentation from other providers also reminds us that “AI” is not one system. OpenAI defines separate bots for making sites available in ChatGPT Search answers, training models, and user-triggered access. ",[148,163,166],{"href":164,"rel":165,"target":154},"https:\u002F\u002Fdevelopers.openai.com\u002Fapi\u002Fdocs\u002Fbots",[152,153],"OpenAI’s crawler documentation"," Perplexity likewise distinguishes the bot it uses to show sites in search results from user-triggered access. ",[148,169,172],{"href":170,"rel":171,"target":154},"https:\u002F\u002Fdocs.perplexity.ai\u002Fdocs\u002Fresources\u002Fperplexity-crawlers",[152,153],"Perplexity’s crawler documentation",[14,174,175],{},"So the sentence “I added a file for AI” is too broad unless it says which company, which product, and which flow it refers to. Appearing in search results, model training, grounding, user-triggered retrieval, and being cited in an answer are not the same outcomes controlled by the same switch.",[14,177,178,179,181],{},"Once we get those distinctions wrong, it becomes very easy to give one ",[32,180,34],{}," file a role it has not earned.",[72,183,185],{"id":184},"the-difference-between-a-good-experiment-and-a-bad-prescription","The difference between a good experiment and a bad prescription",[14,187,188],{},"The fact that an idea is worth testing does not mean it should be recommended to everyone.",[14,190,191],{},"In my implementation, generating alternative content representations fit the existing architecture. If the work amounts to a few templates, a build step, or another representation of data already being produced, the cost may be reasonable. Those files may also become useful for a future documentation, API, or content-processing need—not just for AI.",[14,193,194],{},"On another site, the same work may require a new publishing pipeline, content synchronisation, error tracking, URL maintenance, and editorial discipline. At that point, the question is not, “Is the file harmful?” The question is: “Does this investment make sense even without AI Visibility?”",[14,196,197],{},"If the answer is no, the claimed benefit deserves a higher standard of evidence.",[14,199,200],{},"My practical decision tree is very simple today:",[202,203,204,207,210,213],"ol",{},[25,205,206],{},"Does this alternative representation emerge naturally from the existing information architecture at low cost?",[25,208,209],{},"Would it be useful to another user or product even if no AI system used it?",[25,211,212],{},"Can I keep these copies reliably up to date when the content changes?",[25,214,215],{},"If I see no benefit a year from now, will the investment still look reasonable?",[14,217,218,219,221],{},"If most answers are yes, I will try it. If most are no, I will not prioritise ",[32,220,34],{}," simply because competitors are doing it.",[72,223,225],{"id":224},"another-acronym-tomorrow","Another acronym tomorrow",[14,227,228,229,231],{},"This article is not about banning ",[32,230,34],{}," or dismissing everyone who works with it. I built it too, and I went well beyond a file at the root. I want to draw a line between trying something that is technically meaningful and speaking about it with more certainty than the evidence allows.",[14,233,234,235,237],{},"Today it is ",[32,236,34],{},". Tomorrow it will be another file, schema type, agent standard, or metric. The accessories will change; the way we evaluate a recommendation should not.",[14,239,240],{},"If we want to speak more calmly about AI Visibility, we should ask the same question of every new tactic:",[11,242,243],{},[14,244,245],{},"Is this a behaviour documented by the provider, an observation from research with a transparent methodology, or a hypothesis that still needs to be tested?",[14,247,248],{},"If neither of the first two answers is available, “let’s try it” is sometimes the right answer. But until that experiment produces a result, it should not be presented as non-negotiable.",[14,250,251,253],{},[32,252,34],{}," was never the point. The point is knowing where the evidence ends—and treating certainty beyond that point as a hypothesis rather than a marketing claim.",[72,255,257],{"id":256},"sources","Sources",[22,259,260,266,273,279],{},[25,261,262],{},[148,263,265],{"href":150,"rel":264,"target":154},[152,153],"Google: Guide to optimizing for generative AI features on Google Search",[25,267,268],{},[148,269,272],{"href":270,"rel":271,"target":154},"https:\u002F\u002Fdevelopers.google.com\u002Fsearch\u002Fdocs\u002Fappearance\u002Fai-features",[152,153],"Google: AI features and your website",[25,274,275],{},[148,276,278],{"href":164,"rel":277,"target":154},[152,153],"OpenAI: Overview of OpenAI crawlers",[25,280,281],{},[148,282,284],{"href":170,"rel":283,"target":154},[152,153],"Perplexity: Crawlers",{"title":286,"searchDepth":287,"depth":287,"links":288},"",2,[289,290,291,292,293,294],{"id":74,"depth":287,"text":75},{"id":105,"depth":287,"text":106},{"id":130,"depth":287,"text":131},{"id":184,"depth":287,"text":185},{"id":224,"depth":287,"text":225},{"id":256,"depth":287,"text":257},[296,297],"ai","business",null,"2026-08-23","A closer look at the AI Visibility hype around llms.txt, the information architecture experiment I built, and the difference between a technical hypothesis and a marketing promise.",false,"md","\u002Fimages\u002Fhero\u002Fllms-hype.avif","Essay","en",{},true,"\u002Fllms-txt-was-never-the-point",8,{"title":6,"description":300},"llms-txt-was-never-the-point",[313,314,315,316],"llms-txt","seo","geo","ai-visibility","post","bEE5Ve4jsNlvmuBSVPJsjCd0-vscXkVJN8bOpM1oKCY",{"path":320},"\u002Ftr\u002Fmesele-llms-txt-degil-yegen",{"prev":322,"next":325,"others":328,"lucky":403,"readingTime":309},{"path":323,"title":324},"\u002Fai-visibility-tools-reality-or-a-construct","AI Visibility Tools: Reality or a Construct?",{"path":326,"title":327},"\u002Fdo-you-know-how-dependent-your-company-is-on-ai","Do You Know How Dependent Your Company Is on AI?",[329,332,335,336,337,340,343,346,349,352,355,358,361,364,367,370,373,376,379,382,385,388,391,394,397,400],{"path":330,"title":331},"\u002Fturkeys-first-real-time-mystery-shopping-reporting","Turkey's First Real-Time Mystery Shopping Reporting Platform",{"path":333,"title":334},"\u002Fkeeping-customers-happy-isnt-enough","Keeping Customers Happy Isn’t Enough. You Have to Follow Up.",{"path":323,"title":324},{"path":326,"title":327},{"path":338,"title":339},"\u002Fthe-ai-productivity-baseline-is-moving-faster-than-we-remember","AI Wasn’t Always This Good. We Just Got Used to It.",{"path":341,"title":342},"\u002Fai-made-code-cheap-verification-is-still-expensive","AI Made Code Cheap. Verification Is Still Expensive.",{"path":344,"title":345},"\u002Fthe-threshold-collapsed-to-zero","The Threshold Collapsed to Zero: Building in Public in the Age of AI Cloning",{"path":347,"title":348},"\u002Fbuild-in-public-2-0","Build in Public 2.0: The New Playbook for Sharing Openly in the Age of AI Cloning",{"path":350,"title":351},"\u002Fgoogle-generative-ai-data-ai-citation-timing","AI Citations Run on a Separate Clock: What Google's Generative AI Data Revealed",{"path":353,"title":354},"\u002Fthe-era-of-the-previous-vibe-coder-begins","The Era of the \"Previous Vibe Coder\" Begins: The Invisibility of Clean Code and the Technical Debt Bill of AI",{"path":356,"title":357},"\u002Fone-victory-several-defeats","One Victory, Several Defeats",{"path":359,"title":360},"\u002Fthe-job-ai-wont-take-and-the-five-it-prevents","The Job AI Won't Take — and the Five It Quietly Prevents",{"path":362,"title":363},"\u002Fproductlog-the-platform-i-built-for-myself-first","ProductLog: The Platform I Built for Myself First",{"path":365,"title":366},"\u002Fcomprehension-debt-the-bill-comes-due-alone","Comprehension Debt: The Bill Comes Due Alone",{"path":368,"title":369},"\u002Fthe-new-evrenbal-com-a-new-era-and-blogging-philosophy","The New Evrenbal.com: A New Era and Blogging Philosophy",{"path":371,"title":372},"\u002Fai-visibility-illusion-bing-citation-share","The AI Visibility Illusion: What Bing's Citation Share Data Actually Reveals",{"path":374,"title":375},"\u002F1m-impressions-per-month-0-revenue-a-programmatic-seo-post-mortem","1M Impressions per Month, $0 Revenue: A Programmatic SEO Post-Mortem",{"path":377,"title":378},"\u002Fthe-end-of-coding-or-a-new-renaissance-the-invisible-crisis-of-ai","The End of Coding or a New Renaissance? The Invisible Crisis of AI",{"path":380,"title":381},"\u002Fraising-children-in-the-age-of-artificial-intelligence","Raising Children in the Age of Artificial Intelligence",{"path":383,"title":384},"\u002Fredar-ai-powered-summaries-for-kap-disclosures-and-open-sources","Redar: AI-Powered Summaries for KAP Disclosures and Open Sources",{"path":386,"title":387},"\u002Fpesintaksit-cash-vs-installments-a-turkish-inflation-aware-payment-comparison-tool","Cash or Installments? – The Story Behind PeşinTaksit",{"path":389,"title":390},"\u002Fbeyond-the-bot-lessons-from-building-a-chat-system-for-global-patients","Beyond the bot: lessons from building a chat system for global patients",{"path":392,"title":393},"\u002Fhow-i-built-a-modern-infrastructure-using-open-source-tools-and-the-power-of-cloudflare","How I Built a Modern Infrastructure Using Open Source Tools and the Power of Cloudflare",{"path":395,"title":396},"\u002Ffrom-rules-to-decisions-the-real-time-sales-intelligence-platform-we-built-at-vanity","From Rules to Decisions: The Real-Time Sales Intelligence Platform We Built at Vanity",{"path":398,"title":399},"\u002Fwhy-im-building-rankextension-making-google-search-console-actually-make-sense","Why I’m Building RankExtension: Making Google Search Console Actually Make Sense",{"path":401,"title":402},"\u002Fan-seo-experiment-in-a-low-competition-serp-with-google-maps-and-openai","An SEO Experiment in a Low-Competition SERP with Google Maps and OpenAI",{"path":392,"title":393},[405,407,409,410],{"path":350,"title":351,"date":406},"2026-06-28",{"path":371,"title":372,"date":408},"2026-06-22",{"path":323,"title":324,"date":299},{"path":374,"title":375,"date":411},"2026-06-21",[413,415,416],{"path":359,"title":360,"date":414},"2026-06-25",{"path":344,"title":345,"date":406},{"path":380,"title":381,"date":417},"2025-11-14",1787524295316]