Ask ChatGPT which project management tool works best for a five-person agency, and you’ll often get an answer that reads like it was assembled from a conversation rather than a landing page. That’s because it usually was. A growing share of what AI answer engines cite comes from forums, and Reddit sits near the center of it. People searching for Cory Maki Reddit guidance normally want the same thing my clients want: the mechanism, not the mystique. So let’s walk the actual path a thread takes from posted comment to cited sentence.
A quick definition before we go further. Generative Engine Optimization (GEO) is the practice of getting your brand, product or expertise represented accurately inside AI-generated answers — ChatGPT, Perplexity, Google AI Overviews and the rest. An AI citation is when one of those systems pulls from a source and either links it or paraphrases it into the response. Rankings are a separate game. In AI search, citations are the currency.
Why Reddit keeps showing up in AI answers
Answer engines are built to resolve questions, and Reddit is a decade-plus archive of questions being resolved by people with no incentive to flatter a product. Threads carry structure that models find easy to work with: a specific question in the title, competing answers underneath, replies that challenge weak claims, and visible signals of what the group found useful.
That combination is rare. A vendor blog post asserts. A Reddit thread argues, corrects itself, and lands somewhere. When a model needs to say “most users report X,” a thread with forty comments is a far more defensible source than a page selling the thing. I’ve written more about that dynamic in why AI engines love Reddit and treat community as citation.
Over a decade in reputation and search, I’ve watched the same pattern repeat: the surfaces that feel least “optimizable” end up carrying the most weight, precisely because they’re hard to fake at scale.
The path from post to citation
Here is the sequence as I explain it to founders. It isn’t magic and it isn’t instant.
1. A real question exists and keeps getting asked
Citations start with demand. Somebody posts “What are people using for HIPAA-compliant intake forms?” That phrasing — long, specific, conversational — mirrors how people prompt AI tools. Threads that match high-intent question patterns get retrieved far more often than threads about general opinions.
2. Someone answers with specificity, not a pitch
The comment that gets quoted is almost never the enthusiastic one. It’s the one that says which tool, for what use case, at what team size, with what tradeoff. Named entities and concrete constraints are what make a passage extractable. Vague praise is unquotable — there’s nothing in it to lift.
3. The community reacts, and that reaction becomes signal
Replies agreeing, disagreeing, or adding nuance tell a model that the answer was tested. Upvotes matter, but so does the shape of the discussion: a comment with three substantive replies confirming the point reads as consensus. A comment sitting alone reads as one person’s opinion.
4. The thread gets indexed and stays retrievable
Reddit content is crawled aggressively and surfaces in traditional search results for long-tail questions. That visibility is what puts it in range when a retrieval system goes looking for supporting material. Old threads matter here — a 2022 comment can still be the best available answer to a 2026 prompt.
5. The model synthesizes, then attributes
At answer time, the system pulls passages, blends them with other sources, and produces a response. Sometimes you get a direct link. Sometimes your product name appears in a sentence with no link at all. Both count. Being named inside an answer is LLM visibility, whether or not anyone clicks.
A concrete example
Imagine a SaaS founder whose tool handles contract review for small law firms. A thread appears: “Solo attorney here — is contract review software worth it or should I just use templates?”
The losing move is a comment that says “Check out our tool, it’s great for solos.” It gets downvoted, possibly removed, and contributes nothing extractable.
The winning move is a comment from someone who actually knows the space: templates work fine until you’re handling more than a handful of nonstandard agreements a month; the real cost of review software isn’t the subscription, it’s the time to train it on your clause library; here’s the threshold where the math flips; and yes, disclosure, I build in this category. That comment answers the question, earns replies, and contains a quotable, bounded claim. Months later it’s the kind of passage a model reaches for when someone asks whether solo attorneys need contract software.
That’s the whole game. Not presence — usefulness that happens to be attributable.
The Cory Maki Reddit approach: show up where answers form
In my work with clients — startups, law firms and public figures — the Reddit component is never a campaign. It’s a participation habit with structure around it. Here’s the practical version:
- Find the threads that are already ranking. Search your category’s real questions and note which Reddit threads appear in Google results and in AI Overviews. Those are the pages with retrieval gravity. Adding a genuinely good comment to a live, visible thread beats starting a new one nobody finds.
- Answer the question asked, fully, before mentioning anything you’re attached to. If your comment would still be useful with your product removed, you’re in good shape.
- Disclose your affiliation plainly. Every time. It costs you nothing and it’s the difference between a contribution and a violation.
- Write in extractable units. Short paragraphs. Specific numbers where you legitimately have them. Named tools and named tradeoffs. Clarity and structure are what make content citable — on Reddit no less than on your own site.
- Stay in a handful of subreddits for the long haul. Community authority is cumulative. One account, consistently useful in three relevant places, outperforms scattered activity everywhere.
- Answer follow-ups. The replies under your comment are part of what makes it look settled rather than asserted.
This connects directly to the ARC Method I created for earning AI citations, and it’s the same logic I apply across owned content and earned media in broader Generative Engine Optimization work. Reddit is one surface. The principle — be the most useful, most specific answer available at the moment the answer forms — travels everywhere.
The things that get you removed instead of cited
I’ll be blunt, because this is where most brands wreck themselves. Fake accounts, purchased upvotes, coordinated comment rings, and “organic-looking” agency posting are not aggressive tactics — they’re liabilities. Moderators catch them. Communities remember them. And a removed comment cites nothing, ever.
There’s also a reputational tail. Reddit threads about brands that got caught astroturfing rank well and stay retrievable for years, which means the misconduct becomes the citation. I’ve catalogued the common failure modes in what not to do on Reddit, and the constructive counterpart in building genuine authority on Reddit. Reputation is earned, not bought — that’s not a moral flourish, it’s an accurate description of how these systems assign weight.
If your category already has negative threads with real traction, that’s a separate discipline. It belongs to AI reputation management, and the fix is almost always more accurate information from credible sources, not suppression.
The durable principle
What I’m seeing across AI search is a steady shift in where influence actually sits. For fifteen years the job was to own a ranking position. Now the job is to be the source a model reaches for when it composes an answer — and models reach for places where humans have already done the work of sorting good advice from bad.
That’s the honest summary of Reddit AI search: you don’t optimize your way in, you contribute your way in. Show up in the threads where your category’s questions get settled, answer them better than anyone else has, say who you are, and let the citations follow. It’s slower than a campaign. It’s also the only version that holds up.