TL;DR. When an AI engine recommends a business, it cites its reading: a "best X in [city]" listicle, a Reddit thread, a review hub, a local news piece. Those exact pages are your to-do list. By the end of this project you can run the citations roadmap: pitch the listicle AI already reads, show up on Reddit as yourself, and collect mentions, on a 2-hour-a-week habit.
Watch us map our own AI citations (the receipts)
This is Project 5 of 7 in the AI SEO playbook, and it is the part almost nobody does, which is exactly why it decides the answers.
- Watch us map our own AI citations (the receipts)
- Why are citations the roadmap?
- What does earning one citation do?
- Why is Step 5 where the answers get decided?
- How do you run the citations roadmap, step by step?
- What did the 2026 ranking data change? (the upgrade layer, added July 2026)
- How do you make this your own?
- How do you train someone else on this?
- Where do you share your result?
- Frequently asked questions
Why are citations the roadmap?
Go back to the sources you saved in Project 1. Every one of those links is a page the AI already trusts enough to cite when it answers your buyers. You do not have to guess where authority comes from in your market: the machine showed you its reading list. Getting onto those pages beats publishing more of your own, because the AI is already reading them.
Our operating rule of thumb: at least 40% of your total effort goes to distribution. Earning your way onto other people's pages, not writing more of your own.
What does earning one citation do?
If ChatGPT keeps citing "The 9 Best Recovery Studios in Austin" when your buyers ask, then one placement on that list beats ten new blog posts. That single page feeds every future answer the engine assembles for that question. This is how the incumbents in your Project 1 audit got there: they are on the pages the machine reads, and every month they stay there, the AI gets more sure of its pick. Advanced Recovery Cryo's climb to the first answer in its market (the case study) ran on exactly this loop: reading the citations, then earning them.
Why is Step 5 where the answers get decided?
Because AI engines synthesize from sources, and community sources weigh heavily. Reddit in particular: engines lean on it hard for "who do locals recommend" questions. A helpful, genuine, real-name Reddit presence turns into AI citations for your business, costs nothing, and beats most of what agencies charge for. Astroturfing gets smelled instantly, by Redditors and by models, so the only version that works is the honest one.
Steps 3 and 4 made your OWN house trustworthy and quotable. Step 5 is the away game, and the away game is where recommendations are won.
How do you run the citations roadmap, step by step?
The drill. Two hours this week, then two hours every week:
- Open your Project 1 citation list. Group the saved links: listicles, Reddit threads, review platforms, directories, local press.
- Pitch the listicle AI already trusts. Find the publisher's contact, and send a short, honest note: who you are, why you belong on the list, one receipt (your review count, a result, a differentiator). No template blast; one real email per list.
- Show up on Reddit as yourself. Real name, real business in your bio. Find this week's question in your city's subreddit or your category's community, and write a genuinely useful 6 to 10 sentence answer with no pitch. One natural line of who you are inside the helpful answer is the signal the AI latches onto.
- Feed the review platforms the AI cited, not just the one you like. If Perplexity keeps reading Yelp for your category, your next five happy customers go to Yelp.
- Collect mentions, not just links. A local paper, a chamber page, a partner's site saying your name matters even without a hyperlink. Machines count mentions.
Log each earned citation next to the prompt it feeds. By Project 7 you will watch placements turn into share of voice.
We grouped every source the engines cited in our own Project 1 audit: 119 citations across 69 domains. Google surfaces alone took 30 of them, six times more than any other source, which is exactly why the boring Project 3 came before this one. Seven "best agency" listicles took 17. Our own site was cited 5 times out of 119 in answers about our own market.
The roadmap wrote itself: pitch notes drafted for each of the seven listicles (one real email per list, no template blast), the two unclaimed directories go on this week's list, and the community rep is one genuinely useful Reddit answer a week under Nick's real name. Every earned citation gets logged next to the prompt it feeds, and Project 7 counts what it does to share of voice.
The receipts live in the Project 5 post in our build-in-public thread on X.
What did the 2026 ranking data change? (the upgrade layer, added July 2026)
The 2026 Local Search Ranking Factors survey put a number on what this project teaches. For classic Google Maps rankings, citations carry about 6% of the weight, a decade-long decline. For AI search visibility they carry 13%, and three of the top five AI factors are citation factors. The survey's author put it in one line: in AI SEO, mentions are the new link. (Whitespark, 2026)
And the single most important AI visibility factor of all 187 measured? Presence on expert-curated "best of" lists. Not your domain authority, not your backlinks. Whether a list a human curated says your name. That is this project, and it is why the pitch step above is worth more than it feels like when you are writing one awkward email.
Two upgrades the data adds to the drill:
- Split your citations into two piles and work them differently. STRUCTURED citations (directories, aggregators, review platforms) are a checklist: fix the four data aggregators once, claim the 15 to 25 directories that matter for your industry, keep the data identical everywhere, and refresh quarterly because engines track citation freshness. EARNED citations (listicles, local press, community mentions) are a pitch motion, and they are the pile that decides AI answers. A sync service can rent you the first pile; nothing can automate the second.
- Do not fake the earned pile. In our July 2026 probes, the most-retrieved source in one product category was the market leader's own comparison listicles: they wrote honest "best of" pages, included competitors by name, and now the engines quote them as the category's reference. That play works. The corrupted version does not: Google has begun penalizing self-serving best-of pages that rank their author first with no methodology, and we have watched engines get gamed by PR-blast mirages and self-listicle farms that fall apart under one day of fact-checking. If you publish comparison content, put real competitors in it, show your criteria, and disclose your bias. The transparency is the mechanism, not decoration.
How do you make this your own?
Your upgrade: the directory only your niche knows. Every industry has one or two directories or associations that machines treat as authorities (the industry body, the certification registry, the specialty booking platform). Get listed there before your competitors realize it is being read, and you own a citation they cannot easily copy.
Selling software? The whole roadmap reprioritizes around brand mentions, which beat backlinks 3 to 4x for AI visibility: the B2B SaaS upgrade has the order. Product brand? The surfaces swap to testing sites, gift guides, YouTube reviews, and annual listicle re-inclusion: the eCommerce upgrade ranks them.
How do you train someone else on this?
Run a mention swap with one non-competing local business you genuinely recommend: you write a real recommendation of them where your audiences overlap, they do the same for you, and you teach them the roadmap while you are at it. Two honest mentions, two businesses the machines now connect to their community. Peer, mentee, mentor: three swaps and the habit is permanent.
Where do you share your result?
Post the win when a placement lands: "We are now on [the list] that ChatGPT cites when people ask for the best [category] in [city]. Here is how we earned it." That post is itself a mention, and it teaches the next owner the method. Link back here and tag us; the best placements get featured on the playbook.
Score yourself: being on at least one page the AI already cites for your market is +1 on the playbook's 0 to 5 AI-awareness score. Next: Project 6 is the honest chapter about volume.
Frequently asked questions
How do backlinks work for AI search?
AI engines cite sources when they answer, and they favor pages that are themselves well-referenced. A backlink or mention on a page the AI already reads (a listicle, a review hub, a community thread) puts your name directly into the engine's reading, which is more direct than classic link equity.
Does Reddit marketing actually work?
The honest version does. Engines weigh Reddit heavily for "who do locals recommend" questions, so genuine, useful, real-name answers in the communities your buyers use turn into citations. Anonymous self-promotion gets flagged by Redditors and discounted by models.
What is a local citation?
Any place your business's name, address, and phone appear online: directories, review platforms, maps services, association pages. Citations on the platforms the AI reads for your market, consistent with your NAP from Project 3, tell machines you are an established, real business.
How long until citations show up in AI answers?
Engines re-crawl and re-synthesize constantly, so a strong placement can surface in answers within weeks. The compounding is the point: each citation makes the next answer more likely to include you, which is why this is a weekly habit rather than a campaign.
Two ways to run this playbook. Get your free AI Visibility Score in about 30 seconds: 50 prompts, 3 AI engines, scored live, plus a free 20-minute walkthrough with Matthew. Or get The Playbook and keep doing it yourself.
