Earned Media for AEO: How to Get Named on the Sources AI Already Cites

Every AEO program starts in the same place. You audit your visibility, find you are at zero, and start publishing. Six months later you have forty answer-first articles and your citation share has barely moved, because the engines answering your category's questions are not reading your blog. They are reading a four-year-old Reddit thread, two YouTube reviews, and a listicle on a publisher you have never pitched.

That is not a content quality problem. It is a surface problem, and it needs earned media to fix.

Why your own blog cannot carry AEO alone

The clearest number on this comes from Stacker's distribution research: content sitting on a brand's own site got cited 8 percent of the time, and the same content distributed through third-party outlets got cited 34 percent of the time. Same words, a 325 percent lift in citation rate, purely from where it lived.

The mechanism is not mysterious. An engine assembling an answer needs corroboration. Your page claims your product is the best option for mid-market teams. Every competitor's page claims the same thing about theirs. The model has no way to adjudicate self-description, so it leans on sources that have no stake in the answer. A Reddit commenter who used three tools and explains why two of them broke is worth more to the model than your feature page, because the commenter had nothing to gain.

Similarweb's analysis of UGC citations puts a finer point on it: content that engines cite runs about 20.6 percent entity density against roughly 7 percent for typical web content. Third-party discussion is dense with named things: products, versions, prices, competitors, specific failure modes. Your marketing page is dense with adjectives. Density of named entities is what makes a passage quotable.

This does not mean owned content is wasted. Owned content is what gets quoted once the model has decided you belong in the answer. Earned media is what gets you into the consideration set in the first place. Teams that skip the second half publish into a room where nobody is listening. If you want the owned side running properly alongside this, the answer is to pair the two rather than treat them as separate programs.

The stats contradict each other, and that is the actual lesson

Read four articles about Reddit and AI citations and you will get four incompatible numbers. Profound's analysis reports Reddit at 46.7 percent of Perplexity's top citations, 21 percent of Google AI Overviews, and 11.3 percent of ChatGPT. Similarweb's reads ChatGPT at 5.3 percent. BuzzStream's study found Reddit at about 3 percent of total citations, with ChatGPT using Reddit zero times in their sample.

Most write-ups pick whichever number is most dramatic and tell you to go do Reddit. The honest reading is different: citation share is specific to the engine, the category, and the query type, and no global number predicts yours.

BuzzStream's own data shows why. In their sample, 83 percent of Reddit citations landed on purchase-decision queries, and 77 percent of Reddit's visibility came from generic category searches rather than review keywords. So a study that tests "best CRM for startups" and a study that tests "what is a CRM" will report wildly different Reddit shares, and both will be right about what they measured.

The same holds for YouTube. MaxAEO's analysis of YouTube AI citations puts YouTube at 38.7 percent of Perplexity's video citations and 36.6 percent of Google AI Overviews, but only 4.4 percent of ChatGPT's. If ChatGPT is where your buyers are, a YouTube-first earned media plan is a very expensive way to move a number nobody is watching.

The practical consequence: you cannot copy someone else's earned media plan. You have to read your own citation data first. This is the same reason one engine cites a page while another ignores it: different indexes, different scoring, different source preferences.

Citation is not the same as a mention, and the gap is where money dies

Here is the failure mode nobody prices in.

You spend three months earning a spot in a comparison roundup. An engine cites that roundup. The answer it generates names two competitors and not you. The citation happened. The mention did not. Your dashboard, if it only tracks cited domains, logs this as a win.

A mention means the engine says your name in the answer. A citation means the engine points at a source. They come from different processes and they come apart constantly. BuzzStream's analysis of the two signals found only 23.1 percent of brand mentions were backed by a citation, while 69.9 percent of citations included the brand's name in the response text. The two numbers move independently, which means you can be cited without being named, and named without being cited.

For earned media this distinction is the whole game. A cited Reddit thread that recommends your competitor is worse than no thread, because it is actively teaching the model that your category's answer does not include you. This is the mechanism behind why AI assistants recommend your competitor instead of you.

So the unit of measurement is not "did we get cited on Reddit." It is named citations per priority prompt, by source. Anything coarser will lie to you.

Step 1: Build your category's cited-source list

Before you pitch anyone, find out who the engines are already reading.

Take 25 to 50 prompts that reflect how your buyers actually ask, weighted toward decision-stage phrasing: category questions, "alternatives to," head-to-head comparisons, problem-first phrasing, and a few branded ones. Run them across the engines your buyers use. Log every cited URL, its parent domain, and crucially, whether your brand was named in the answer text.

Sort the parent domains by frequency. That list is your earned media target list, and it will not look like the global top-10 lists in blog posts. Ours, for the AEO category, is led by YouTube and includes a mix of vendor blogs, practitioner sites, HubSpot, Reddit, and Medium. Yours will be different, and that is the point. If you want this run continuously rather than by hand, measuring citation share across LLMs covers the method in depth.

Do this before spending a dollar. A target list built from someone else's category is how teams end up with a Reddit strategy in a category where the engines never cite Reddit.

Step 2: Rank targets by how you can legitimately show up

Not every cited domain is winnable, and they are not winnable the same way. Sort your list into three buckets.

Open surfaces. Reddit, Quora, communities. Anyone can participate. The constraint is not access, it is credibility, and it is enforced by humans who are very good at spotting marketing.

Contributor surfaces. YouTube, Medium, LinkedIn, your own owned-but-syndicated content. You control the artifact completely. The constraint is that nobody is obligated to watch it, and unwatched content still gets cited, which changes the calculus a lot.

Gated surfaces. Publisher roundups, editorial coverage, review platforms. You need someone to say yes. Slowest, highest trust, most durable.

The sequencing mistake is starting with gated surfaces because they feel prestigious. Start with contributor surfaces, because they are the only ones where you control both the timeline and the content, and because citation does not require popularity. MaxAEO's data found 40.8 percent of cited YouTube videos had under 1,000 views. The engine is not reading your view count. It is reading your transcript.

Step 3: Earn the mention on each surface

Reddit. The subreddits engines cite most have hostile anti-promotion norms and long community memory. Dropping a product link gets you deleted, and a ban is permanent in a way a bad ad never is. The version that works is slow: a real account, months of substantive answers in threads where you happen to know the answer, disclosure when you have a stake, and patience for the payoff, which is other people naming you unprompted. That third-party mention is worth more than anything you can post yourself, and it is the only Reddit outcome that survives the model's skepticism.

If you cannot commit to that timeline, do not do Reddit. A half-hearted Reddit presence is not neutral, it is a liability.

YouTube. The most under-priced surface on the list, because the citation is driven by text you fully control. MaxAEO's cited-video sample averaged 334-word descriptions and roughly 19-word titles, with long-form capturing about 94 percent of YouTube's AI citations against 5.7 percent for Shorts. Treat the description as a source page, not a caption. Name the product, the version, the alternatives, the specific situation. Publish the transcript. A video with no transcript is invisible to the engine no matter how good it is.

Publishers and roundups. Pitch inclusion, not coverage. "Add us to your existing comparison" is a far easier yes than "write about us," and for citation purposes it is worth the same or more, because roundups are exactly the shape of content an engine reaches for on a shortlist query. Bring the reason: the specific segment you win, with evidence.

Structured feeds. Everything above is about pages you do not own. The counterpart is making your own surfaces trivially machine-readable so that when an engine follows a citation back to you, it finds a clean answer. Our AI Feed Engine handles the ingestion side of this, and the free llms.txt generator covers the basic file if you want to start there for nothing.

Step 4: Measure citation lift, not placements

Placements are an activity metric. Citation lift is the outcome.

You already built the baseline in step 1. Now re-run the same prompt set, unchanged, on a weekly cadence, and log named citations per prompt by source. When a placement lands, watch the window after it. Pickup speed differs by engine: Perplexity tends to reflect new earned media within days, while ChatGPT and Google AI Overviews generally take one to four weeks, so a re-audit two to four weeks after a placement is the reasonable checkpoint.

Attribution here is correlational and you should say so out loud. You are matching lift on specific prompts to placements that landed in the same window. That is weaker than a controlled experiment and much stronger than counting logos. The discipline that makes it credible is holding the prompt set fixed. The moment you edit prompts mid-quarter, your trend line means nothing, and every change after that is unfalsifiable.

What good looks like, roughly:

Our own FastTrackr AI case study shows how this compounds once the third-party surfaces start carrying the brand name. And if you are ready to run the loop continuously instead of quarterly, pricing is here.

What to do when the cited thread names your competitor

You will find these. A high-frequency thread that engines cite constantly, recommending someone else. The instinct is to reply and correct the record from a brand account. Do not.

The move is to make the thread's conclusion stale. Threads get cited because they are the best available answer to that question. Publish a better one somewhere the engines already read, aimed at the exact question the thread answers, with the specificity the thread has and your marketing pages lack. Then earn a genuine mention in a newer discussion. Engines re-crawl. Consensus shifts. It is slow, and it is the only version that holds.

The wrong move, brand account arguing in the replies, gets screenshotted and becomes a second cited thread about how your company argues with customers.