Lesson 03·GEO Tactics·~10 min·why → tactics
Where AI actually looks
You fixed your words in Lesson 2. But here's the twist: when AI recommends a business, most of what it quotes isn't your website at all. It's what everyone else says about you.
In ~10 minutes you'll be able to
- Explain why AI quotes other people's pages when it recommends a business.
- Read the off-site "consensus signal" AI engines lean on.
- Audit where you're mentioned — and where you're invisible.
Short on time? Watch the 60-second version.
The whole idea, up front:
When AI answers "best [thing] near me," your website is only one part of the evidence — often a small part. Across AI answers, the large majority of citations go to earned, third-party sources (Muck Rack: ~84% earned media; brand-owned content barely registers), not your own site.[1] AI tends to recommend a business when independent sources agree about it — a complete Google Business Profile, recent specific reviews, consistent listings, authentic community mentions, and credible local press. Make those tell the same true story and you become the safe thing to recommend.
01Your website is necessary, but not enough
This is the counter-intuitive part. Your site still matters — it's where AI verifies the basics — but it's rarely where the recommendation comes from. Muck Rack's analysis of 25 million+ AI-cited links found ~84% of citations go to earned media (independent third-party coverage), with paid and brand-owned content barely registering.[1] And brand mentions predict AI visibility far better than backlinks: Ahrefs' 75,000-brand study found branded web mentions correlated 0.664 with AI Overview visibility versus 0.218 for backlinks — about 3× stronger. (Backlinks still help; they're just no longer the main lever.)[3]
One caveat to hold throughout: "AI" isn't one engine. ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews cite different source types, and the same business can surface differently depending on the exact question, the user's location, and what's available to retrieve.[6]
Where the citations come from
share of AI citations · Muck Rack Generative Pulse 2026 (25M+ links)
02The mechanism: the consensus signal
AI systems appear to reward consistency across sources: the same business name, location, category, services, and reputation showing up in the places they can retrieve. When your Google listing, your reviews, a Reddit thread, and a local "best of" article all describe you the same way, the engine has enough corroboration to recommend you. When sources disagree — or only your own website is talking you up — it stays cautious.[2]
That's the whole game: be the business that independent sources agree about. One analysis found a brand praised across four or more independent forums was about 2.8× more likely to be named by ChatGPT than one only talked up on its own site.[2] The exact multiplier shifts depending on who's measuring — but the direction (more independent agreement → more likely cited) is consistent across studies.
03The five places to be
For a local/small business, these are the off-site signals AI leans on most — roughly in order of leverage. The percentages are one firm's estimates of AI-visibility factors (directional, not official Google or OpenAI weights); Google's own local guidance is simpler — relevance, distance, and prominence, supported by a complete, verified profile.[4]
04Consensus Builder — watch AI's confidence climb
Start with a business that only has its own website. Toggle on each off-site signal and watch how likely AI is to recommend it. This is the off-site sibling of Lesson 2's Rewrite Lab.
Interactive · your feedback loop
Consensus Builder
Click each source to add it to your off-site footprint. This is a confidence simulator to show relative leverage — not a measured score.
Tip: click them off again to see how visibility drops when a source goes missing or inconsistent.
05Check your understanding
2 quick calls
Click an answer for instant feedback.
Q1Someone asks ChatGPT "best plumber in Austin." What most determines whether you're named?
Q2Fastest way to find which subreddits AI trusts for your category?
06The honest caveat
What this can and can't do
You influence off-site signals — you don't control them. You can't write your own Reddit reviews or command a newspaper. The move is to earn the consensus: do work worth talking about, then make it easy to find and consistent across platforms.
Don't fake it. AI cites negative sentiment about as often as positive (~6% vs ~5%), so manufactured praise backfires, and review/forum manipulation is increasingly detected and penalized.[5] And the data is noisy: when Columbia Journalism Review tested eight AI search tools across 1,600 queries, they were wrong on more than 60% — often confidently.[7] Citation shares also swing fast: Reddit's share on ChatGPT dropped from ~60% to ~10% in six weeks after a single change.[6] So diversify across these five sources rather than betting on one, and treat this as reputation work that compounds over months.
Sources
- Muck Rack, Generative Pulse 2026 (25M+ AI-cited links across ChatGPT/Claude/Gemini) — ~84% of AI citations are earned media (held at 82–89% over three quarters); paid 0.3%; brand-owned content barely registers. ↩
- Evertune, How AI systems choose which brands to cite — a brand praised across 4+ independent forums was ~2.8× more likely to appear in ChatGPT than one mentioned only on its own site (Evertune attributes this figure to Clearscope). Direction is consistent across studies; the exact multiplier varies. ↩
- Ahrefs, AI Overview Brand Visibility (75,000 brands) — branded web mentions correlated 0.664 with AI Overview visibility vs 0.218 for backlinks (~3× stronger); the top factors are all off-site. Backlinks remain a weak positive signal, not a negative one. ↩
- Advice Local, 2026 Local Search Ranking Factors — directional estimates (on-page ~24%, reviews ~16%, citations ~13%, GBP ~12%), not official ranking factors. Google's own local guidance: relevance, distance, prominence + a complete, verified profile. ↩
- Profound, The data on Reddit & AI search (4B+ citations) — AI picks 3–5 subreddits per query; pulls negative sentiment (~6.1%) about as often as positive (~5%); authentic help beats promotion. ↩
- 5WPR, AI Platform Citation Source Index 2026 (680M citations) — Reddit ~40% (most-cited single domain, volatile: 60%→10% in 6 weeks); top-15 domains = 68%; different engines cite different source types. ↩
- Columbia Journalism Review (Tow Center), "We compared eight AI search engines…" — across 1,600 queries the tools were wrong on >60%, often with confidence. AI citation accuracy is poor. ↩
— end of lesson 3 —