Chapter 11 — Key Takeaways

A one-page reference. If you remember nothing else: search is not only text, and the way to win the visual surfaces is to describe what you actually have — clearly, honestly, and fast.

The one idea

Images, video, and visual search are large, under-contested search surfaces that reward the same honest work as the rest of SEO — describe your media truthfully (for the screen-reader user first, Google second), keep it fast, and match what the searcher wants — and punish the shortcuts, because you cannot keyword-stuff a photograph.

How Google understands an image (and who controls what)

Signal Who controls it Weight
Alt text You High for image search; describe, don't stuff
Surrounding text / caption You Real — context disambiguates the image
Filename You Minor but free (anode-rod.jpg > IMG_4821.jpg)
Structured data You Explicit facts; enables rich results (Ch 18)
Page context / quality You Google prefers images from good pages
Machine vision (pixels, OCR) Google Powerful, not tunable — give it a clear subject
EXIF geotag/keywords You Ranking myth. Only IPTC licensing metadata is real

Rules of thumb

  • Alt text is for people first. Describe the image in context, concisely; empty alt="" for decorative images; never stuff keywords. The image-search benefit rides along.
  • Images are where speed dies. Heavy hero image → hurts LCP; unsized image → hurts CLS. Use modern formats (WebP/AVIF), size correctly, set dimensions. (Full weighting: Chapter 16.)
  • Never lazy-load your LCP image. Lazy-load below the fold; eager-load the top. Standard lazy loading does not hide images from Google — only broken, interaction-gated loading does.
  • For video, eligibility ≠ entitlement. VideoObject markup and key moments make a video eligible for rich results; Google never guarantees them.
  • YouTube is a search engine. Optimize title, description + timestamped chapters, a custom thumbnail, accurate captions, and — above all — watch time. Don't obsess over tags.
  • You can't "optimize for Lens." Visual search matches pixels and facts, not keywords. Publish clear, single-subject, indexed images with accurate structured data.
  • Sitemaps aid discovery, not ranking. Image/video sitemaps help Google find media (especially via JS/CSS); they never lift it.

Bust these myths

Myth Reality
"Geotag EXIF to rank locally" No evidence Google uses it; local signals are Ch 25 (GBP, NAP, content)
"Keyword-stuff every alt attribute" Hostile to screen readers, a spam pattern, and it describes nothing
"Lazy loading hides images from Google" Standard lazy loading is supported; only broken versions hide
"There's a secret Google Lens trick" No dial, no tag — clarity + accurate data is the whole game
"A video thumbnail is guaranteed traffic" Google redefined eligibility in 2023; rich results are rented

The evidence, honestly

  • Alt text is confirmed by Google to help understand and rank images; it is a weak lever for web ranking. Its airtight case is accessibility (WCAG).
  • "YouTube is the #2 search engine" is directionally true (huge, intent-rich demand), not a cleanly measured rank — don't stake anything on the exact number.
  • Core Web Vitals is a modest tiebreaker, not a magic bullet (Chapter 16).
  • IPTC licensing metadata is the one documented embedded-metadata use — it earns the "Licensable" badge, not rankings.

What you can do today

Run a free alt-text/accessibility audit (WAVE or Lighthouse) on your five most important pages: fix missing and stuffed alt text so each meaningful image is described, and give decorative images an empty alt="". Then open PageSpeed Insights on your heaviest page and note whether the LCP element is a bloated image. Two free tools, one afternoon, real improvement.

Where this goes next

Chapter 12 turns from making media to managing content over its lifecycle — auditing, updating, merging, and (counter-intuitively) deleting — the truth that publishing less can grow traffic more. Later, Chapter 16 measures the image-speed work, and Chapter 18 implements the VideoObject markup sketched here.