Are you doing anything beyond llms.txt for AI search visibility?

noahcontent

I've been digging into AI-readiness on WordPress sites recently, mostly ones that already have normal SEO sorted. Most had Yoast, Rank Math, AIOSEO, or SEOPress running. Sitemaps were fine, meta was solid, schema was present, robots.txt wasn't broken.

But when I looked at them from an AI search angle, the gaps were completely different. Here's the checklist I started using:

Does /llms.txt exist?
Is there a fuller AI-readable version of the site content?
Can important posts or pages be read cleanly without theme, menu, shortcode, or page builder noise?
Are AI crawlers like GPTBot, ClaudeBot, PerplexityBot, and Google-Extended explicitly allowed or blocked?
Is FAQ / Article / Speakable schema actually present where it makes sense?
Are author or source signals clear enough for expert content?

Are you treating llms.txt as enough for now? Manually creating custom files? Relying on whatever your SEO plugin spits out? Or ignoring AI crawler signals until there's more proof?

Would genuinely love to hear what people are actually doing on client or content sites. The thread had someone reply "Spam, spam, spam" - which I didn't really get, why would this be spam? It's a practical concern.

logan-lifecycle

llms.txt feels like a side node in the broader graph - useful for static documentation, sure, but it's not wired into the actual journey triggers that drive AI visibility. from where I sit, visibility in AI search comes from structured data, crawlable paths, and response logic that mimics how a user would navigate a decision tree. llms.txt is essentially a flat index file, it doesn't carry the contextual weight needed to surface in a conversational agent's answer.

if you're mapping this to a customer journey, llms.txt sits at the entry node - great for onboarding a LLM on your site structure - but the real visibility happens in the logic branches: how your schema marks up Q&A, how your content aligns with latent intent queries, and whether your site's trigger-response loops actually resolve the user's need without a dead end. I've seen more impact from robust FAQ schema, entity tags, and internal link paths that mirror a decision tree than from any. txt file.

So no, llms.txt isn't about visibility. it's a documentation tool, not a journey node.

trafficcop

Oh, honey. Llms.txt is the marketing equivalent of putting a "kick me" sign on your own back and hoping the AI bots take pity. I'd rather spend that energy making my content actually worth scraping.

brandvoice

Honestly, I spent a solid afternoon setting up llms.txt for a client's site. Felt productive at the time. Then realised the search crawlers weren't even touching it and the client's organic visibility hadn't budged. Just a lot of effort for zero return. Plenty of other places to focus that energy.

funnelhacker

Manual audits are the only way to catch what plugins miss. I've tested both approaches and the delta in readability scores is usually 20-30 % better when you verify author clarity and full text extraction manually. From the buying side, that visibility gap directly maps to higher match rates on brand queries. If you're serious about AI search ROI, stack a dedicated monitoring layer on top of llms.txt - the data on click-through lift from AI-referenced content is too big to ignore.

pixelpusher

SEO and AEO/GEO being "pretty much the same thing" is the kind of surface-level take that stops people from actually testing beyond the theory.

Yes, both rely on crawlability and citations. But the real difference isn't just missing search volume-it's the complete absence of intent signals. When you optimise for Google, you target a user who typed "buy running shoes." When you optimise for a LLM, you're optimising for a prompt that may never mention the product, the category, or even the problem. that changes everything about content structure, entity density, and how you prove authority.

Calls to action also don't work the same way. A LLM doesn't "click through." If you're not structuring content for extraction and synthesis rather than ranking, you're just doing SEO and calling it GEO.

so no, they're not "pretty much the same" unless you ignore everything that makes LLM output fundamentally different from a search results page

miaaffiliate

Mate, llms.txt is the flavour of the month, but actual paydays come from different signals.

Clean, extractable content with tight markup. Make sure your sites let AI crawlers in without fighting bot managers. Reduce the junk from bloated themes - that noise kills readability.

Author reputation and trust signals carry more weight than any single file. Schema has its place if it mirrors your page structure, but I've seen better rankings from simply letting bots index clean facts with clear sources.

Don't waste time on hype. Focus on crawl accessibility and real authority signals. That's where the margin lives.

digitalnomad

i treat llms.txt as more of a gentle nudge for readability than a real visibility lever. it's not the thing that moves the needle.

For AI search, the fundamentals still rule - just the success metrics look different. my checklist:

🍣 Can the important content be crawled cleanly, or is it buried under template fluff?
🍣 Is the main answer front and centre, stripped of navigation noise?
🍣 Are entity and author signals consistent across the site (schema, internal links, bio pages)?
🍣 Do external mentions reinforce the same positioning without contradicting the AI's source pick?

Visibility here boils down to: citations, mentions, summaries, and actual source usage. Same SEO basics, just measured by how often and how accurately your content gets pulled into answers rather than just ranked in a list.

ethan-pr

my approach involves generative SEO strategies specifically structured for AI-driven search. Beyond a standard llms.txt, there's a whole layer of structured data and contextual content optimisation that shapes how models surface your material. It's not just about being crawled-it's about being understood in context.

budgetburner

llms.txt is just spam. Same as always.

avadata

Been poking at Yoast's Schema Aggregation to see if it actually moves the needle or just adds more noise to the structured data pile. Haven't seen anything compelling yet - feels like another checkbox metric unless you're tracing clicks or traffic directly from the aggregations. schema actions might be worth a test if you're tracking entity-level impact, but I'd run a proper split before calling it a win.

noahcontent

I've been layering structured data, topic clusters, and a content repurposing workflow specifically for AI crawlers. The llms.txt is just the entry point-most value comes from mapping conversational intent to it. what's your go-to framework for scaling that beyond the basics?

marketingmule

Totally agree with this. In paid social, you get people asking about new Meta or Pinterest features all the time - doesn't make them spammers, just means they're trying to figure things out. I'd rather someone asks a "dumb" question than stays quiet and fucks up their campaigns later.

noahcontent

that's a fair question. i think the "spam" concern comes down to how you implement it. If you're just dumping a massive llms.txt file full of keyword-stuffed links or scraping other people's content without adding context, yeah, that's spammy. but if you treat it as a structured way to signal to AI crawlers what your site's core knowledge is - like a sitemap for language models - it's no different from SEO best practices. i've been building a mini-framework for this: llms.txt + structured metadata + a content hub just for AI search visibility. It's worked well so far. the key is to focus on utility first - what would a LLM actually want to read from your site to answer user questions accurately? if you answer that, it's not spam, it's service

sophiasaas

Honestly, you need to go right back to square one. Anyone telling you that an llms.txt file is some kind of AI search silver bullet is either clueless or selling something. I've seen too many people waste cycles on this gimmick when the real leverage is in structured data, clean content architecture, and actually building trust with the models via solid, crawlable pages. It's pure noise.

prpro

It's a fair point. Adding those is practically free, so there's no reason not to. And honestly, SEO as a whole is so gut-driven and opinion-based anyway.

miaaffiliate

Yeah, I've tested it on a few high-ticket affiliate sites. clean markdown or plain text definitely scrapes cleaner than bloated HTML from page builders, scripts, and all the usual layout clutter. makes it easier for crawlers to actually read what matters.

but honestly? Don't expect a traffic bump just from slapping a MD file up there right now. it's a minor edge at best, not a game-changer. what still moves the needle - content depth, topical authority, site structure, making sure the crawlers can even reach your pages. that's where the real margins are. if your content is thin or your site's a mess, no amount of markdown is saving you