14 prompts across 3 LLMs - zero citations

paperclick

Ran a test last Thursday - buyer-intent prompts through ChatGPT, Claude, and Gemini. Things like "best tools for [category]" and "what should a Series B use for [workflow]." Fourteen prompts total. We appeared in zero.

Main competitor showed up in eleven of fourteen. Eleven. Screenshot every response, built a deck because my CEO didn't believe me. He does now.

What kills me is we outrank them organically for roughly 80% of those terms. Two years of SEO budget and apparently none of it transfers when someone asks an LLM instead of Googling.

Also found our knowledge base hasn't been updated since March. Cool. Different fire.

Has anyone figured out how to get cited by these things? Or are we all just watching brands disappear? My entire visibility strategy has a hole in it and I don't even know which department owns this mess.

emailwiz

I remember when we first jumped on the AI bandwagon for our content generation - thought we'd cracked the code. Set up a whole pipeline: every category prompt fed to three different models, refined, cross-referenced. same result: zero appearances. Felt like shouting into three separate wells and hearing nothing but our own echo. Honestly, digital marketing now is just that: three different voids you're screaming into, hoping one of them gets annoyed enough to toss you back a response.

ppcnerd

Two years of SEO spend, grinding for that page one slot, and now the game's changed completely. It's like building a billboard on a highway that everyone's stopped driving on. We spent all that time optimising for Google's algorithm, and the user's moved on to a conversational interface that doesn't even look at our carefully crafted meta descriptions.

the macro trend is pretty clear: search is becoming an API call, not a browsing session. Your ranking work isn't worthless-it's just that the floor's been pulled out from under the old model. We've been playing chess on a board that's being reassembled mid-game. The real lesson here is that channel dependency is a fragile bet. I've seen it in video ads too-platform changes can obliterate a year of optimisation overnight.

It stings, but it's a reminder that the only durable skill is understanding the user's intent, not just the platform's rules. SEO as we knew it isn't dead-it's just been demoted to a background player. The new front page is whatever the LLM decides to spit out first.

sleepermode

I'm wrestling with the exact same question internally. If the LLMs are ingesting your content but never surfacing your brand, where does ownership of that break-down actually sit? Is it SEO (because it's about discoverability and citation signals), content (because the material exists), brand (because it's about perception), or an entirely new discipline that doesn't have a seat at the table yet?

No one in our org is volunteering to own it either - you're instantly accountable for something no one knows how to measure or attribute. We don't even have a job family that covers "LLM citation performance."

chatbox

We've run into the same wall with clients. organic rankings don't just carry over to the AI chatbots - the models pull from totally different sources: comparison pages, documentation, forums, listicles, that kind of thing.

what cracked it for us was treating this as source mapping rather than SEO reporting. I've been using a tool to track which prompts surface competitors, what citations appear, and where the gaps are for each client. then you build the missing citable content around the sources those models already trust.

If you get that content live and structured properly (schema, clear factual statements, authoritative backlinks), the models start picking it up. It's slow - think months, not weeks - but it does work.

coopplayer

We ran the exact same experiment with Semrush, and honestly, the difference between LLM results and traditional SEO rankings is absolutely mad. It's like your brand exists in two parallel universes - one where you're invisible and one where you're everywhere. And seeing that big fat zero? Brutal. lol

dusting

tried that experiment myself. Felt like trying to pin down a stable signal in shifting sands - rephrase the prompt even slightly and the results scatter. Asked for "best tool" versus "what to use" and got nearly half the brands different. makes you question how reliable any of these systems are for serious category analysis.

weekendhustle

I've seen the exact same pattern play out across a few B2B SaaS clients we've worked with. It's not an SEO problem - it's a citation source problem. ChatGPT and Gemini pull from a completely different pool than Google's organic results. For any tool category, the big three are Wikipedia, G2, and Forbes. That's where the AI is scanning. If your competitor has a solid Wikipedia page, ranks in the right G2 category, and gets a mention in a Forbes roundup, they'll show up in AI answers even if you're wiping the floor with them on every Google keyword.

The stale knowledge base thing you noticed? That's huge. AI models weigh recency heavily. If your product docs, pricing, or comparison content haven't been touched in months, the model treats your info as less reliable than a competitor who pushed updates last week. For one of our clients, the fastest fix wasn't creating more content - it was a structured refresh of existing pages with current pricing, new features, and clear use-case descriptions. Made a noticeable difference within weeks.

Structure matters too. AI loves tables, lists, and direct answers over narrative paragraphs. If your competitor has a comparison table on their site with pros and cons, that table gets extracted verbatim. If your site relies on long-form blog posts, the model has to summarise and often drops your brand name in the process. We saw that happen repeatedly with one brand - their name just disappeared from AI summaries even though they were the better product.

So start with the citation sources that actually feed AI visibility: clean up your G2 profile and make sure you're in the right categories. If you qualify for Wikipedia, that's the single highest-leverage thing you can do across every AI platform. Then audit your own site for extractable content - comparison tables, FAQ sections with natural language questions, clear pricing pages. That's where the real gap is, not in search rankings.

growthgeek

this comparison is lazy. paid placement inside a LLM is fundamentally different from search ads-there's no query volume guarantee, no CPC model that makes sense, and the training data is already locked. people keep saying "it's like early AdWords" but that only works if you assume LLMs will let you bid on terms. i doubt they will, they want to own the output, not auction it off. Have you actually seen any provider signal a paid tier for brand inclusion, or is this just wishful thinking based on old habits?

trendspotter

Still, LLM traffic remains a mere shadow of Google's blazing sun.

metricsmadness

That's a tough spot to be in-feels like setting up shop in a ghost town and wondering why no one walks through the door. Could you share what category you're in and which LLMs you tested? might be able to trace where the signal gets lost-whether it's crawl gaps, content structure, or just the way these models summarise your niche. Happy to take a look if you want.

paperclick

been there. Ranked #1 on Google but zero AI mentions. trust signals are the new SEO.

ethan-pr

There's no quick fix to get featured in LLM responses, and the whole industry is still figuring out what works. The general direction is shorter, answer-style content placed higher on the page, paired with a technically clean site-fast load times, proper schema markup where it makes sense. That's the short version, anyway.

If this is going to be a focus (and it really should be), I'd strongly recommend a citation tracker. Otherwise you're manually hunting through prompts across platforms over time, which is unsustainable. I've used Citation Hawk and Otterly AI for this-both have their merits depending on your budget and workflow.

ctrjunkie

God, that's brutal. Especially when you know you're absolutely smashing it in organic - that sting is real. The knowledge base piece is probably a massive factor. These models seem to grab whatever's freshest and most linked across the web, so if you're not in those specific sources, you're invisible.

I've been talking to a lot of marketers in my community who've hit the same wall. We've had to treat LLM optimisation like a completely different beast now. It's not really an SEO team problem anymore - it's more content and product teams working together to get mentioned in the right contexts. Different game entirely.

marketingmule

I've been saying this since GPT-3.5 and getting eye rolls from leadership every time. The LLM layer is eating the informational query space, but nobody wanted to hear it because budgets were still flowing. Now the traffic graphs dip and suddenly executives want a deck explaining why. It's the same story as when featured snippets ate clicks-except this time the answer doesn't even link back to you. Anyone else's leadership still pretending this isn't happening?

wordweaver

oh honey, you ran all those prompts and got zilch? colour me shocked. LLM recommendations are basically a slot machine with worse odds. teams burn days trying to game the algorithm when they could just.... talk to people who are actually asking for help.

someone on here looking for a tool rec is already holding their wallet out. that's a warm lead. chasing an AI hallucination? that's just free labour for OpenAI. go find the real conversations instead

backlinker

Makes sense. Ran the same audit through Semrush - the domains with heavy LLM citations all had solid schema, especially FAQ and HowTo structured data. Ours was pretty thin on the ground, so zero visibility wasn't unexpected. For anyone in a similar boat:

{
 "@context": "https://schema.org",
 "@type": "FAQPage",
 "mainEntity": [{
 "@type": "Question",
 "name": "Your core question",
 "acceptedAnswer": {
 "@type": "Answer",
 "text": "Your concise answer."
 }
 }]
}

That alone won't fix everything, but it's a baseline the LLMs seem to index. without it, you're basically invisible to their extraction pipelines

maya-b2b

Honestly, figuring out whether an LLM cited you or just mentioned your category feels like an attribution nightmare piled on top of the one we already have. Seen this with clients too and it's yet another layer of complexity nobody asked for lol.