Short answer: We ran 50 small-business queries through ChatGPT, Perplexity, and Google AI Mode in July 2026. Perplexity cited an SMB primary site 64% of the time, AI Mode 41%, and ChatGPT 27%. The businesses that got cited weren't necessarily ranking #1 in Google — they shared four structural signals (AI crawler access, schema, llms.txt, and a verified entity) far more often than the ones that didn't. This post walks through the data and the playbook.
Why We Ran the Study
Every client conversation in 2026 ends in the same question: "Are we getting cited?" Rank data answers a question that no longer maps to traffic on its own. We wanted a clean, reproducible baseline that shows which AI engines actually quote small businesses, and what those businesses have in common.
So we picked 50 queries a real SMB would care about — "best HVAC company in Calgary," "what is fractional CMO pricing," "how to choose a Vancouver web design agency," "Shopify vs WooCommerce for a home goods brand," and 46 more — and ran them through the three most-used AI search surfaces today.
The Headline Numbers
- Perplexity cited an SMB primary website in 64% of queries.
- Google AI Mode cited an SMB primary website in 41% of queries.
- ChatGPT (GPT-5 with web) cited an SMB primary website in 27% of queries.
- Across all three engines, directories and review aggregators (Yelp, Houzz, Clutch, G2, BBB) accounted for a third or more of cited URLs.
- Editorial sources (Forbes, trade pubs, local newspapers) appeared in nearly every ChatGPT answer and roughly half of AI Mode answers.
Translation: getting cited is harder on ChatGPT, easier on Perplexity, and middling on Google AI Mode. But across all three, the businesses that did get cited were not random.
The Four Signals Shared by Cited Sites
1. AI crawlers were allowed
Of the SMBs cited across all three engines, 96% allowed GPTBot, PerplexityBot, ClaudeBot, and Google-Extended in their robots.txt. Of the SMBs that ranked well organically but never got cited, 61% blocked at least one of these crawlers — usually inherited from a CDN, security plugin, or default WordPress template. This is still the cheapest fix with the biggest payoff.
2. Structured data was present and correct
Cited sites carried Organization schema on the homepage (with sameAs pointing at LinkedIn and Wikidata), FAQPage schema on service pages, and Article schema with honest dateModified on blog content. Uncited sites either had no schema or had broken schema (we found Google Tag Manager-injected JSON-LD with syntax errors on three of the uncited sites).
3. An llms.txt file existed at the root
Only 18 of the 50 query winners had a published llms.txt, but those that did were cited more than twice as often per query as ones without. Perplexity, in particular, appears to weight it heavily. If you don't have one yet, our llms.txt guide walks through a copy-paste template you can ship in 30 minutes.
4. The brand was a verified entity
Cited businesses overwhelmingly had a Wikidata entry, a claimed Knowledge Panel, and a fully completed Google Business Profile with current photos and review velocity. This is the structural work behind GEO and AEO — it's also the highest-leverage one-time task most established SMBs still haven't done.
What Didn't Predict Citation
A few things we expected to matter, didn't:
- Domain Authority / Domain Rating. No meaningful correlation. A DR-22 plumbing site in Surrey was cited by all three engines for a local query a DR-71 directory missed.
- Word count. The cited pages skewed shorter, not longer, when the question was direct. The median cited paragraph was 47 words.
- Pure ranking position. Several cited sources ranked positions 4–9 organically. The AI is reading the page, not just the rank.
What This Means for Your Strategy
The takeaway is not "abandon SEO." It's that SEO without GEO/AEO is now half the job. The organic visibility you already have is what makes you discoverable; the structural signals above are what make you quotable. The businesses winning AI citations in mid-2026 are doing both — and the gap between them and competitors who are still rank-only is widening month over month.
If you want a faster path to the four-signal checklist, our GEO services page describes the full implementation work, and a free 30-minute consultation will get you a query-level look at where your business currently shows up across AI Mode, ChatGPT, and Perplexity.
What to Read Next
- Google Made AI Mode the Default → The bigger context behind why citation share is now a top-line KPI.
- The 9-Signal Citation Checklist → The full implementation guide that maps to the patterns in this study.
- The Complete llms.txt Guide → The 30-minute task most cited sites had done.
Find out where you stand
Get a query-level look at your AI citation share
Book a free 30-minute consultation. We'll run your top commercial queries through Google AI Mode, ChatGPT, and Perplexity — live — and map the four-signal gaps holding you back.
Book My Free Consultation →30 minutes · No obligation · Vancouver-based, serving North America

