Off-Page AEO: Earn the Citations That Actually Win

The biggest source of AI citations is not your website, it is the people writing about your category. When we logged all 1,237 citations the four major engines produced across 47 buyer questions, 40% pointed at independent third-party listicles and only 34% at the vendors’ own sites. The rest went to media, Reddit, review sites, and YouTube. Every one of those surfaces is off your domain, and every one of them can be earned.

This is the most important chapter in the playbook. It gives you eight plays, in order of impact, that get your brand onto the places AI already trusts: review sites, directories, best-of lists, Reddit, press. (We call these places “surfaces” throughout; a surface is simply anywhere AI reads and cites.) You will finish with a 20-surface target list for your own brand and your first five pitches drafted.

One honest note before the plays: this is the slowest chapter in the book and the one most worth doing. On-page fixes ship in an afternoon. Earning a place in someone else’s list takes weeks. But a placement, once earned, keeps answering for you.

Why off-page beats on-page right now#

Ask the engines who to buy from and they mostly do not read the brands. They read the writing about the brands.

Bar chart. Share of all 1,237 citations, by source type: Independent listicles 40%, Vendors' own sites 34%, Media & SaaS blogs 15%, Reddit & forums 4%, Review sites (G2 etc.) 3%, YouTube 3%.

Reuse this chart

Free to reuse with attribution. Copy the embed code, or save the image.

Download PNG Copy embed code

<figure>
  <img src="https://visibly.so/figures/citation-2026-07-source-types-ch7.png" alt="Bar chart. Share of all 1,237 citations, by source type: Independent listicles 40%, Vendors' own sites 34%, Media & SaaS blogs 15%, Reddit & forums 4%, Review sites (G2 etc.) 3%, YouTube 3%." width="720" height="288" style="max-width:100%;height:auto" />
  <figcaption>Where AI search citations actually go by <a href="https://visibly.so/playbook/off-page-aeo/">Visibly</a></figcaption>
</figure>

The external research points the same direction. Search Engine Land’s 2026 review found Reddit, YouTube, and LinkedIn the most-cited sources across AI engines. Search Influence’s analysis found review-site profiles roughly triple the likelihood of a ChatGPT citation. A network-level study reported by openPR put numbered listicles near 63% of AI citations. And one 2026 industry analysis estimates the median AI-citation lift from earned media at around +239%, the largest of any tactic it measured. Those are third-party numbers, so treat the exact figures as directional. The direction, though, is unanimous.

There is a second, quieter reason off-page wins: memory. ChatGPT skipped live web search on 29 of its 47 answers in our study and answered from what it already knows. No on-page fix reaches a model that is not reading pages. What reaches it is years of other people writing about you: the entity records, reviews, and coverage the model learned from.

The five kinds of places that earn citations#

Every play in this chapter gets you onto one of five kinds of places. Together they cover about two-thirds of all citations we measured.

The five kinds of places that earn AI citations, converging into one AI answer. Entity records (Wikidata, Crunchbase, LinkedIn, Google Business Profile, Play 19), review platforms (G2, Capterra, TrustRadius, Trustpilot, Play 20; g2.com was the #4 most-cited source in the study), directories (Product Hunt, Futurepedia, AlternativeTo, Play 21), communities (Reddit, YouTube, Quora, Play 23; reddit.com was the #1 most-cited source), and lists and press (Zapier, TechRadar, Forbes, Plays 22 and 24; listicles took 40% of all citations).

Who actually sits at the top of those places today? Straight from the study’s scoreboard:

Bar chart. Citations across all four engines, top 8 of 713 domains: reddit.com 43, youtube.com 32, zapier.com 31, g2.com 19, salesforce.com 13, techradar.com 12, monday.com 11, forbes.com 10.

Reddit, YouTube, Zapier, G2. Not one of them is a brand marketing site. Those are the kinds of places this chapter helps you get into.

The plays#

Eight plays, ordered by impact for a brand starting from little or nothing. Every effort and time-to-impact figure is an estimate, not a measurement. AI answers change from run to run, so each play ends with a check you can repeat.

Play 19

AnyDo this first

Claim your entity across Wikidata, Crunchbase, and LinkedIn

Make your brand resolvable, so models answering from memory describe the real company instead of guessing.

Why it works

ChatGPT answered 29 of 47 buyer questions from memory with no live search in our study, so what the model already knows about you decides those answers. Entity records are the most direct write path into that knowledge. (Visibly AI Citation Study, checked 2026-07-10)

Steps

1. Write one canonical boilerplate: legal name, one-sentence description, category, founding year, founders, HQ, and domain. Every profile uses exactly this.
2. Complete your Crunchbase profile with the boilerplate, logo, and links.
3. Create a Wikidata item: instance of (your company type), official website, industry, founders. The notability bar is low; the resolution value is high.
4. Rewrite your LinkedIn company page About section with the same boilerplate.
5. Claim your Google Business Profile, even without a storefront.
6. Publish the same description on your site's about page and in your llms.txt, so on-site and off-site records agree.

Go here Wikidata: new item Crunchbase LinkedIn company page Google Business Profile

Tools All free: Wikidata, Crunchbase, LinkedIn, Google Business Profile

Effort About half a day (estimate, not measured)

Time to impact Weeks to months; entity answers usually move before citations do (estimate)

Done when: All profiles live with identical boilerplate, and your about page matches them.

Verify it worked: Ask ChatGPT and Perplexity 'What is [your brand] ([your domain])?' and check the answer describes the real company. Re-ask monthly.

Common failure mode: Drafting the profiles and never publishing them. We did exactly this for three weeks, and the model kept hallucinating our company in the meantime.

Play 20

B2B SaaSAgency

Get your first 10 reviews on G2 and Capterra

Put your brand on the review surfaces AI engines read and cite when they compare tools.

Why it works

g2.com was the #4 most-cited source in our study (19 citations) and one of only 5 domains cited by all four engines. Third-party analysis by Search Influence found review-site profiles roughly triple the likelihood of a ChatGPT citation. (Search Influence, checked 2026-07-10)

Steps

1. Create your G2 listing: right category, complete profile, real screenshots, honest feature checklist.
2. Repeat on Capterra (and TrustRadius if you sell mid-market or up).
3. List your 20 happiest users or design partners, and email each one personally with a direct review link and a specific mention of the work you did together.
4. Target 10 verified reviews in 60 days; follow up once, politely.
5. Respond to every review, including critical ones. AI reads the whole page.

Go here G2 for vendors Capterra for vendors TrustRadius for vendors

Tools G2 and Capterra listings are free; the asks are time

Effort 2-3 hours of setup, then about 30 minutes a week (estimate)

Time to impact Review-effect observations in third-party tests land within 1-2 months (estimate)

Done when: Listing live in the correct category with 10+ verified reviews.

Verify it worked: Search your category on G2 and confirm you appear. Then ask an assistant 'best [your category] tools' monthly and watch whether review-site content carries your name.

Common failure mode: Incentivizing reviews outside platform rules, or sending one generic blast. Ten personal asks convert better than a hundred automated ones, and fake reviews get platforms delisted.

Play 21

Any

Submit to the directories that already cite your competitors

Replicate the cheapest, most repeatable pattern in your rivals' backlink profiles.

Why it works

Directories are the most repeated pattern in competitor link profiles. When we mined our own direct rivals, 16 niche directories linked to two or more of them before we had submitted to a single one. Links a rival earned are proven, replicable routes.

Steps

1. Pull the referring domains of 3-4 direct competitors (any backlink tool's free tier works).
2. Filter to directories and tool catalogs in your niche; ignore generic link farms.
3. Build one reusable submission kit: logo files, 50, 100, and 160-character descriptions, category tags, three screenshots, founder bio.
4. Submit to the top 15, starting with any that link two or more rivals.
5. Log the live URL of every accepted listing; a submission without a live URL does not count.

Go here Product Hunt Futurepedia: submit There's An AI For That AlternativeTo Ahrefs free backlink checker

Tools Ahrefs or Semrush free tier for the mining; a spreadsheet for the log

Effort One hour for the kit, then about 10 minutes per directory (estimate)

Time to impact Listings land in days to weeks; the authority compounds over months (estimate)

Done when: 15 live directory listings, each URL logged.

Verify it worked: Load every listing URL, then re-check your referring-domain count a month later.

Common failure mode: Blasting 50 generic AI directories nobody reads. The 15 directories your competitors' backlinks prove matter beat any volume play.

Play 23

Any

Show up on Reddit before you need it

Earn presence in the single most-cited source in AI search, authentically enough to keep it.

Why it works

reddit.com was the most-cited individual source in our study (43 citations), and Search Engine Land's 2026 review found Reddit the #1 cited source across AI engines. In our data, community citations concentrate almost entirely on Perplexity, which produced 65 of the 75 Reddit and YouTube citations. (Search Engine Land, checked 2026-07-14)

Steps

1. Warm one or two real accounts for 2-4 weeks with genuine participation and zero links.
2. Map the 5-8 subreddits where your buyers actually ask questions (search your category terms plus 'recommend', 'vs', 'worth it').
3. Run a weekly search for new threads asking your money questions.
4. Answer 3-5 threads a week with real substance. Mention your product only when it directly answers the question, and disclose that you work on it.
5. Prioritize older threads that already rank in Google for your queries; those are what AI re-cites.

Go here Reddit search

Tools Free: Reddit search plus a site:reddit.com Google query

Effort 2-3 hours a week, ongoing (estimate)

Time to impact Perplexity first, since community citations concentrate there; months on other engines (estimate)

Done when: A four-week streak of substantive answers in your mapped subreddits with no removals.

Verify it worked: Ask Perplexity your money prompts monthly and check whether threads you participated in appear as sources.

Common failure mode: Link-dropping from a fresh account. One ban loses the surface entirely, and the communities AI reads most are exactly the ones that ban fastest.

Play 24

B2B SaaSEcommerceAgency

Turn journalist requests into earned citations

Convert expert commentary into the press mentions that carry the highest measured citation lift.

Why it works

Earned media is the highest-lift off-page source in third-party measurement: one 2026 industry analysis estimates the median AI-citation lift from earned media at around +239%. Treat the exact number as directional; the ranking is consistent across sources. (Paid AI Search encyclopedia, ch. 14, checked 2026-07-14)

Steps

1. Sign up for Featured, Qwoted, or Connectively and set alerts for your category's topics.
2. Answer three requests a week. Lead with specific data only you have; journalists lift numbers, not adjectives.
3. Keep answers under 150 words with a one-line bio that includes your domain.
4. Log every placement, linked or unlinked.
5. Once a month, politely ask for a link on any unlinked brand mention.

Go here Featured Qwoted Connectively

Tools Featured and Qwoted free tiers to start

Effort About an hour per answer, three a week (estimate)

Time to impact First placements in weeks; authority compounds over months (estimate)

Done when: Your first earned placement is live, logged with its URL.

Verify it worked: Track referring domains monthly and watch which placements the engines start citing when you re-run your prompts.

Common failure mode: Generic expert-quote answers. Without a specific number or a firsthand observation, your quote is interchangeable and gets cut.

Play 25

B2B SaaSEcommerce

Replicate your competitors' most winnable links

Turn rival backlink profiles into a routed, prioritized to-do list instead of a vanity report.

Why it works

Every link a rival earned is a proven route someone will say yes to again. When we benchmarked our own category, direct rivals held between 821 and 3,052 referring domains before we had earned our first, and the overlap concentrated in a knowable set of directories, roundups, newsletters, and podcasts.

Steps

1. Export the top referring domains for your 3-4 closest competitors.
2. Intersect them: any domain linking to two or more rivals goes to the top of the sheet.
3. Classify every domain by route: directory, roundup, guest post, podcast, news, community.
4. Send each row to its matching play (directories to Play 21, roundups to Play 22, and so on).
5. Work the top 20 by authority times replicability; re-pull the data quarterly.

Go here Ahrefs Semrush

Tools Ahrefs or Semrush trial, or a backlink API; a spreadsheet

Effort About half a day for the mining; execution rides the other plays (estimate)

Time to impact Follows whichever play each target routes into (estimate)

Done when: A deduped target sheet exists with a route assigned to every domain.

Verify it worked: Each month, count how many of your new referring domains came off the sheet.

Common failure mode: Copying junk. A rival's spammy links are not targets; keep a domain only if you would genuinely want the mention.

Play 26

Any

Run a listicle radar so you never miss a new list

Catch new best-of lists in your category within days, while their authors are still shaping them.

Why it works

The list ecosystem moves fast. Between two of our own weekly citation checks in July 2026, roughly 20 tools we were not tracking entered the lists AI cites for our category. New lists are easiest to join while they are still fresh and their authors are still responsive.

Steps

1. Set Google Alerts for 'best [your category] tools' and 'best [your category] software'.
2. Once a month, re-ask the engines your money prompts and diff the cited domains against last month's log.
3. Pitch the author of any new list within days, using the Play 22 pitch format.
4. Keep every placement and pitch in the same tracker, with live URLs.

Go here Google Alerts

Tools Google Alerts (free) plus the engines

Effort About 30 minutes a week (estimate)

Time to impact Fresh lists tend to respond faster than established ones (estimate)

Done when: The radar is running and you have pitched at least one list younger than a month.

Verify it worked: After each placement goes live, re-run the matching prompt and check the citation.

Common failure mode: Treating your target list as static. The set of lists AI cites changes monthly; a one-time pitch batch decays fast.

Where this failed for us#

We drafted the entity profiles from Play 19 (Wikidata, Crunchbase, G2) on June 19, 2026, and then did not publish them for more than three weeks while we shipped content instead. In that window our weekly checks kept showing ChatGPT hallucinating a different company when asked “What is Visibly?”. The lesson we paid for: preparation is not the work. The send is the work. The drafts earned nothing until they were live, and the surfaces we had not touched kept answering without us. This chapter is written in the order we should have executed it.

Real examples: who wins off-page today#

thedigitalprojectmanager.com What happened: it was one of only 5 domains, out of 713, cited by all four engines in our study. Why it works: it is an independent review site focused on one category, updated all the time, with named authors, clear criteria, and real evaluations. That is exactly the kind of neutral source models reach for. What to do: find the equivalent independent site in your category and make sure they know you exist (Play 22). Source: Visibly AI Citation Study, 2026-07-10.

zapier.com What happened: 31 citations, the #3 source in the study, across many categories at once. Why it works: hundreds of structured “best apps” pages with clear criteria, kept fresh, each one answering a buyer question directly. What to do: if you write comparisons, write them the same way: clear criteria, honest verdicts, current dates. Source: same study, 2026-07-10.

g2.com What happened: 19 citations, #4 in the study, and cited by all four engines. Why it works: verified reviews from many users are the closest thing the web has to neutral proof, and models treat them that way. What to do: your G2 profile is not a badge, it is a page AI reads. Complete it and get reviews on it (Play 20). Source: same study, 2026-07-10.

Do it yourself, or have it run for you#

Everything above is doable in-house with a spreadsheet and 4-6 focused hours a week; that is exactly how this playbook is written, and the worksheet above is the checklist. If nobody on the team has those hours, the honest alternative is not skipping off-page (it is the biggest lever in the data), it is having someone run the same plays for you. That is what Visibly Managed does: the entity work, review motion, pitching, and community participation above, executed and reported against your own citation scoreboard. Either way, the plays are the plays.

For what the winners’ pages themselves look like, the teardowns chapter pulls them apart one by one. And if you have not classified your category yet, start with the citation-pattern chapter, because three industries in eight should run this chapter second, not first.