Skip to content

David Shapiro / Reviewed November 14, 2026

AI repeats the room.

The sources around your buyer belong in the growth plan. Here is how I find them and give them something worth using.

The sources around the buyer
“The room” is my model for the sources around a buying decision. It is not a map of an AI product’s internal process.

Here’s the simplest way I know to explain how AI picks brands: it repeats the room.

Look at the sources around a buying question: publications, reviews, lists, forums, videos. Those voices help shape what a buyer hears about your category. They also give AI search products information to retrieve and use.

Build the useful answer on your site. Then earn support from sources that reach the same buyer.

So the job isn’t “do GEO.” The job is to shape the room.

Your website is not the main voice in the room

This surprises most marketing teams.

One analysis of 500+ commercial queries, across six categories, found that 85% of brand mentions in AI answers came from other sites. Only 13.2% came from the brand’s own domain (AirOps, a content software vendor).

And nearly 90% of those outside mentions came from three formats: lists, comparisons, and reviews. That’s the room.

Your own site still matters — it’s where models go to check prices and facts. Those outside voices deserve a place in the plan.

Step 1: Map the room

You can’t shape a room you haven’t seen. And every category’s room is different.

  1. Write 10–20 prompts your buyers actually ask. “Best [category] for [use case].” “[You] vs [competitor].” “Is [category] worth it?”
  2. Run each one several times, in each engine you care about. Answers vary run to run, so one screenshot won’t do.
  3. Record every source cited. Not just the footnotes — the URLs the engine opened, where you can see them.
  4. Mark each source: does it name competitors? Does it name you?

The sources that name your competitors and skip you are gaps to investigate. When eight competitors are in an article and you’re not, you probably weren’t rejected. You were forgotten. Check the audience, the criteria, and your product fit before you contact anyone.

Make the list useful. Group repeated sources, check which audiences they serve, and prioritize the ones you can make a real contribution to. Twenty relationships is a plan. Two thousand backlinks is a fantasy.

Step 2: Pick your words

Be clear about the problem you want people to connect with your brand.

Try this with a brand you know well. Ask an AI tool what it knows about the brand, and it will list what the brand does. Then ask which brands to choose for one of those things. Record whether the brand appears in both answers.

Being known isn’t enough. You have to be known for something, in words your buyers understand.

So before you pitch anyone:

  • Ask the models: “What do you know about [brand]?” and “What do you know about [brand] for [topic]?”
  • Flip it: “Which brands are best for [attribute]?”
  • Pick three phrases you want AI to attach to your brand. Use them to keep the story coherent. Adapt the language to the audience; do not force keywords into every mention.

Record useful mentions with and without links. The audience, context, and accuracy matter alongside the URL.

Step 3: Win the few URLs that matter

Not all rooms are run by the same people. Find out who runs yours.

Lists that AI keeps citing. In one category, the #1 spot on a cited list was associated with 13.4 percentage points higher AI visibility, in a study of 200,000 AI responses across 8 engines (Peec AI, a tool vendor). The effect was strongest in newer, fragmented categories, and the gains flatten after a few placements. Go deep on the lists that matter, not wide on every list.

Review sites. G2 and Yelp are both in the ten most-cited domains across ChatGPT, Google AI Mode, Gemini, Perplexity, and AI Overviews, in an analysis of 30 million citations (Peec AI, a tool vendor). Find the review sites the engines cite for your category. Keep legitimate profiles accurate and invite honest reviews from customers. Do not select only happy customers or invent reviews.

Publications. Find the 10–15 outlets AI cites most for your space. Pitch them something they want to run: original data, a sharp market view, a live demo. Not “please mention us.”

Smaller voices. A specialist newsletter, trade publication, or expert community can reach a highly relevant audience. Check the audience and the actual contribution you can make before chasing the largest traffic number.

People, not logos. In B2B, put your experts in front: LinkedIn articles from real people, podcast interviews, conference talks. A podcast transcript ties your name to your topic in words that both people and models can read.

How to win them: relationships, not blasts. Use a small list you understand and a specific reason for each contact. Give publishers what they need to add you — proof, data, a sample entry written in their format.

Step 4: Feed the room with one big idea

Start with one idea strong enough to support several useful formats.

One proprietary data set — a survey, an internal benchmark, an original analysis — can become a report, a data visual, a press angle, a dozen social posts, a podcast talking point, updates to relevant existing pages, and useful regional analysis where the underlying data supports it. One effort, many entrances to the room.

It also gives journalists and AI something they can’t get anywhere else. In Orbit Media’s 2026 survey of 1,042 content marketers, just 14% said their blog delivers strong results — the lowest in twelve years. The tactics that go with strong results include influencer collaboration, original research, keyword research, and a formal editing process. All of them are in decline (Orbit Media, Andy Crestodina). The effort is the moat.

Plan distribution before production. Before anyone writes a word, decide where it will travel: which outlets, which creators, which communities, which newsletters. If you can’t name the room it’s going to, don’t make it.

Step 5: Tell the same story everywhere

Keep important brand facts consistent and easy to check. A detailed false story can travel too.

That cuts both ways. In one experiment, models preferred a detailed story that a Reddit thread, a Medium post, and a blog agreed on over a vague official answer (Ahrefs).

  • Write one or two sentences that say what you do. Use them on every profile: LinkedIn, directories, review sites, your founder’s bio, press releases.
  • Fix wrong facts at the source. Find the page that the AI cites for the wrong claim, and ask the owner for a correction. Then publish the right answer, in plain words, on a page you control.
  • Publish an official facts page. What’s true, what’s not, current prices.

What not to do

  • Don’t fake independence. No undisclosed paid posts, fake users, or accounts that hide who they work for. Disclose paid relationships and sponsorships. Build something you can stand behind.
  • Do not build the whole plan around one platform. Be on Reddit, YouTube, or LinkedIn because the audience and contribution fit. Recheck the sources as products change.
  • Don’t blast. Reporters drown in AI-written pitches. A small list you know by name beats a big list you bought.
  • Don’t count syndicated copies as wins. Syndication is for reach. Reporting every copy as a new “citation” is how agencies pad reports.

The short version

  1. Map the room: the sources AI actually cites for your buyers’ prompts.
  2. Pick three phrases you want AI to attach to your brand.
  3. Prioritize the relevant sources in your own observation panel.
  4. Feed the room with one big, original idea at a time.
  5. Tell the same story everywhere, and fix wrong facts where they start.

That’s earned reach with a clear job: give useful expertise a route to the people deciding what to buy.

Request your free manual AI visibility snapshot →

Questions people ask

Is this just PR? It connects PR, SEO, content, and distribution around a specific buying decision. The observation panel helps choose sources. Your audience and the quality of your contribution help decide which ones deserve the work.

How fast does it work? A live-search product can find new coverage after it becomes accessible and retrievable. Publication takes its own time. Future model training follows schedules your team does not control. I plan the work and measure the observations; I do not promise a date for a recommendation.

Do we need Wikipedia? Start with Wikipedia’s notability and conflict-of-interest rules. Then assess its relevance to the audience and category. It is not a shortcut every brand needs.

A note on the evidence

AirOps sells content software. Peec AI sells visibility tracking. Ahrefs sells SEO software. Orbit Media sells website and marketing services. Their research can be useful; their commercial interest belongs beside the findings.

The list-placement comparisons and Orbit survey are observational. They do not establish that a placement or tactic caused the reported outcome. Ahrefs tested a fabricated brand; that experiment is not a universal rule for factual accuracy.

The room is my planning model. Sources can inform live retrieval now and may appear in future training later. A new mention does not instantly change a deployed model’s learned knowledge. There is no fixed number of sources that controls every category.

Explore the source-checked AI search statistics →

Sources

Sources reviewed November 14, 2026.

Try this with your team.

  1. Choose one question a real buyer asks before building a shortlist.
  2. Repeat it in the consumer products that matter to your audience.
  3. Record brands, visible sources, dates, missing answers, and failed checks.
  4. Choose one information, access, or evidence problem to investigate. Record the change before repeating the checks.

Take it to your next meeting

Get the Room Map.

A CSV template and printable instructions. Record the sources, gaps, owners, and next actions behind your buyer questions.

This saves your request and notifies David. No newsletter signup. No analytics or session recording on this page.

Work with David Shapiro

See where your brand stands.

Five buyer questions. Three AI products. Your brand and two competitors. A free manual review with three priorities.

For established brands considering engagements from $5,000/month.

Have a broader project? Tell me about it →
Get your free snapshot