Images stored in CDN-backed object store, not base64 or direct server uploads

Base64-encoded images in JSON-LD or direct-to-server uploads bloat HTML payloads by 30-40%, blow up your origin bandwidth bill, and destroy Core Web Vitals. LCP spikes past 4 seconds on mobile, browsers cannot cache per-image, and you lose responsive format negotiation (WebP, AVIF) that CDNs handle automatically. Google downranks slow pages, paid traffic converts worse, and every listing view costs you compute you should not be paying for.

Move all listing images to a CDN-backed object store and persist only the resulting URL in your database. Use S3 + CloudFront, Cloudinary, or Vercel Image Optimization so browsers get cached, format-negotiated responses instead of inline base64. Migrate existing rows with a backfill job that uploads each embedded payload to the bucket and rewrites the field.

const { Location } = await s3.upload({
  Bucket: 'directory-listing-images',
  Key: `listings/${listing.id}/${filename}`,
  Body: imageBuffer,
  ContentType: mimeType,
  CacheControl: 'public, max-age=31536000, immutable'
}).promise()
await db.listing.update({ where: { id: listing.id }, data: { imageUrl: Location } })