Quick answer: AI engines understand media through its text scaffolding: descriptive file names, specific alt text, captions, surrounding copy, transcripts for video, and ImageObject/VideoObject schema. Optimising media for AI search means making every visual asset fully legible as text – because that text is what gets retrieved, summarised and cited.
How AI engines process your media
Multimodal models can increasingly interpret images directly, but retrieval pipelines still run on text. When Perplexity or an AI Overview surfaces an image or video, selection is driven by the asset’s textual context – the page it lives on, its markup and its metadata. Weak text scaffolding makes strong media invisible.
Image optimisation checklist
File names that describe: geo-audit-dashboard-indexgraph.png, not IMG_4021.png.
Alt text with specifics: “Bar chart showing citation share across four AI engines, June 2026” – facts, not “chart image”.
Captions for key images: visible captions are extracted more reliably than any other image text.
Compression and dimensions: slow images drag page speed, which suppresses crawl coverage (why speed matters).
ImageObject schema on originals – diagrams, charts and product shots you want attributed.
Video optimisation checklist
Full transcripts on the page. The single biggest win: your spoken expertise becomes citable text.
Chapters with descriptive titles – each chapter is an extractable answer unit.
VideoObject schema with name, description, upload date and thumbnail.
A text summary above the embed answering the video’s core question, so engines (and skimmers) get the answer without pressing play.
Original visuals as citation magnets
Original charts and diagrams get cited and reused with attribution – especially when they visualise your original statistics. A branded chart carrying a unique data point is one of the most durable link and mention earners in GEO.
FAQ
Do AI engines cite images directly?
Perplexity and AI Overviews display images from cited pages; the citation goes to the page. Optimised media raises the page’s overall selection odds – track results per prompt with IndexGraph.ai.
Need your media library made AI-legible? Adexorb Technologies handles it inside every GEO program.