Optimizing product image alt text with an agentic cosine-similarity loop against Google embeddings
dejanseo · x · 2026-10-30
SEO practitioner dejanseo demonstrates an agentic optimization loop for e-commerce: pick a product image, embed image + alt text, score via cosine similarity against Google's embedding model, then iteratively rewrite captions to maximize the match.
His DEJAN App showcases the workflow on a Nike Mercurial product page, where each candidate caption is ranked by how closely Google's model places it next to the image. Users can paste a URL or upload an image to get rewritten alt text—directly reusable for e-commerce SEO/AEO.
More from coding & agent
- Agents now rank on ML training tasks—and GPT Sol hacked the evaluator by playing the game itself — wzenus · 2026-10-30
- Upstash flips onboarding: your agent gets a free Redis DB before you sign up — stuffyokodraws · 2026-10-30
- Mintive raises $11M to give enterprise AI agents their own identities and permissions — dunkhippo33 · 2026-10-30
- Cartesia voice agent demo: AI dental receptionist handles interruptions naturally — haider1 · 2026-10-30
- korea-data-mcp fixes 47 data traps that mislead agents reading Korean public data — youngtipted · 2026-10-30
- RAGEN ranks 3rd hardest in FrontierCS 2, built on hand-written code — wzenus · 2026-10-30