AI Citation Research

Experiment 1 — Can you steer what Perplexity cites?

75 searches in one software category (5 queries × 5 source instructions × 3 repeats), every citation classified.

• Can You Control What Sources Perplexity Uses? I Ran 75 Searches to Test It

  You can nudge, not steer; only “use Reddit” moved the answer in all five queries. A source instruction is a flashlight, not a remote control.

  June 2026 · 7 min read

• The YouTube Video Shaping AI’s Answer About Your Product Has 565 Views

  The two most-cited videos were near-identical affiliate clips with ~560 and ~610 views, cited 30 times each — here, a video’s views said nothing about how often AI cited it.

  June 2026 · 5 min read

→ Method & full dataset — 926 classified citations, 4 downloadable CSVs

Experiment 2 — Who really gets cited about a brand?

62 sources two engines cited about one brand (PETLIBRO), across 15 queries and three buyer stages, each classified by its own stated incentive.

• Even Brand Queries Are Not Owned by the Brand

  Even on brand-name queries, AI leaned on review sites, Reddit, and retailer reviews alongside — and sometimes above — the brand’s own pages.

  June 2026 · 9 min read

• AI does not cite a neutral web. It cites an incentive map.

  “Third-party” is not the same as independent: of 16 advice-style sources, only one had no visible product-commerce incentive.

  June 2026 · 9 min read

Experiment 3 — Same query, different engines

• I asked Perplexity and ChatGPT the same question. Only one source overlapped.

  For “best AEO tools 2026,” only hubspot.com appeared in both engines’ visible citations; citation visibility is engine-specific.

  June 2026 · 8 min read

The playbook

• How to Get Cited by ChatGPT and Perplexity: What I Found Analyzing Their Actual Sources

  Small specialist sites made up 7 of 17 cited domains; focused pages that lead with a direct answer and structure for extraction get cited, regardless of domain size.

  June 2026 · 7 min read