Perplexity's Photon post reveals how its ranking engine picks pages — a GEO playbook
gaganghotra_ · x · 2026-10-15
Perplexity's engineering team published a Photon write-up (Sept 24, 2026) detailing its retrieval and ranking engine: p99 latency dropped from 800ms to 65ms, but the bigger story for SEOs/GEOs is the disclosed ranking logic. Each index shard retrieves candidates, runs initial ranking, then a second-stage ranking before merging into the shortlist the answer model sees. First-pass filtering behaves like a classic search engine — rewarding pages that use the right keywords in the right fields without stuffing, ordered by a page-level usefulness signal. Author Ryan Jones calls it a published ranking playbook, more actionable than the Google leak.
More from Apps
- Developer takes OpenAI's new Decisions API grocery shopping in real-world test — prd_008 · 2026-10-15
- User Fills 7-Page Reimbursement PDF From Emails in 30 Seconds With GPT 6.1 — soumitrashukla9 · 2026-10-15
- Wabi AI assistant flags Wealthsimple phishing scam before user even opens it — joonasvirtanen · 2026-10-15
- ChapterPal's PDF-to-Markdown converter goes standalone with LaTeX math and tables — burkov · 2026-10-15
- HeyGen Voice turns any book into an audiobook via ChatGPT MCP integration — HeyGen · 2026-10-15
- Try Gemini Argon in gr.workflow with your OpenRouter key, plus a 3-minute typewriter build — erhannah · 2026-10-15