Chapter 11: Image, Video, and Visual Search SEO

"A search engine will happily rank a photograph it cannot see and a video it never watched — but only if someone bothered to tell it, in words and in markup, what the picture and the film are actually of." — constructed for this chapter, in the spirit of the craft

Overview

Here is a question most site owners never ask, and it is worth more than it sounds: when a person searches with a picture instead of a phrase, or scans a results page for a video before they read a word — is your business anywhere in the answer?

For a decade, "SEO" and "text" were nearly synonyms. You researched keywords, you wrote pages, you earned links to pages. But search stopped being ten blue links of text a long time ago — you saw that in Chapter 1's SERP anatomy. A modern Search Engine Results Page (SERP) is threaded with image packs, video carousels, and thumbnails, and an enormous amount of searching now happens somewhere that is not a text box at all: inside Google Images, inside YouTube — routinely described as the world's second-largest search destination — and, increasingly, through a camera, with Google Lens, where the query is a photograph and the answer is whatever Google can recognize in it. This is a huge surface of search demand, and most sites leave it almost entirely on the table. They upload a phone photo called IMG_4821.jpg with no description, embed a video Google never indexes, and never once consider that a customer might point their camera at a leaking valve instead of typing "what is this part."

That neglect is the opportunity. Image and video search reward the ordinary discipline this book keeps teaching — describe what you actually have, make it fast and legible to a machine, and match what the searcher wants — and they punish the shortcuts, because there is no clever way to keyword-stuff a photograph. The work here is unusually honest: a well-described, well-performing, genuinely useful image or video is a real asset that competitors who are chasing tricks will simply never build.

This chapter covers all three visual surfaces — images, video, and visual search — and it does so with two threads running underneath. The first is accessibility: the single most important image-SEO technique, alt text, exists first to serve people who cannot see the image, and treating it as a keyword-dumping ground is both bad practice and bad ethics. The second is the technical foundation (theme 4): images are the number-one cause of slow pages, so this chapter overlaps hard with Core Web Vitals, and video and visual search lean on structured data — two topics whose full treatment lives later in the book, which we will point to precisely rather than re-teach.

In this chapter, you will learn to:

• Read an image the way Google does — filename, alt text, page context, structured data, and machine vision — and know which signals you can actually influence.
• Write alt text that is genuinely useful to a screen-reader user and, as a by-product, to Google Images.
• Spot the image-performance failures that quietly wreck a page's speed and layout stability.
• Plan on-site video that Google can index, and understand what VideoObject markup and key moments add.
• Treat YouTube as the search engine it is, and optimize a video honestly for it.
• Take visual search and Google Lens seriously, and prepare your images for a camera-first world.
• Help Google discover your images and videos with sitemaps, and earn the "Licensable" badge with metadata.

Learning Paths

Everyone should read §11.1 and §11.2 — how Google understands images, and how to describe them without lying to a screen reader — because they are the foundation the rest builds on. 🏪 Local Business: weight §11.2 (alt text on your service and location images), §11.5 (a how-to video is the highest-trust content a trades business can make), and §11.6 (Lens can identify a storefront or a part). 📝 Content Creator: §11.5 (YouTube as a search engine) is your chapter within the chapter, with §11.4 (key moments) close behind. 🛒 E-Commerce: §11.6 (visual search is reshaping product discovery) and §11.7 (image sitemaps and licensing) are where your money is; product-image quality is a ranking asset, not a nicety. 🔧 Developer: §11.3 (performance, formats, lazy loading) and §11.4 (VideoObject implementation) and §11.7 (sitemaps) are your sections — with a running eye on the Chapter 16 overlap. 📊 Strategist: read the whole thing as diversification — visual search is traffic that does not flow through the ten blue links, and §11.5–11.6 is where you find it.

11.1 How Google understands an image

Start with the fundamental problem, because it explains every technique that follows. A photograph is, to a computer, a grid of colored dots. Nothing in that grid says "this is a corroded water-heater anode rod" or "this is a licensed electrician in Cedar Hills." A human glances and knows; a machine has to infer. For most of the web's history, Google could barely see the picture at all, so it leaned almost entirely on the words around the image to guess what the image was. That has changed — Google's machine-vision capabilities are now genuinely strong — but the older signals still carry enormous weight, and, crucially, they are the ones you control.

We can define the umbrella term now. Image SEO is the practice of helping search engines discover, understand, and surface your images — both as results in their own right (in Google Images and the image packs woven into ordinary results) and as signals that help the page they sit on. It is a small discipline with a few high-leverage moves, and the first move is understanding what Google actually reads.

FIGURE 11.1 — "How Google understands one image"                     [schematic — not to scale]

           ┌──────────────────────────────────────────────────────┐
   YOU     │  FILENAME        anode-rod-corrosion.jpg  (descriptive)│  signals you control
  CONTROL  │  ALT TEXT        "A corroded water-heater anode rod…"  │  ────────────────────▶
   THESE   │  CAPTION / NEARBY the paragraph and heading around it  │        Google's
           │  STRUCTURED DATA  ImageObject / Product / VideoObject  │      understanding
           │  PAGE CONTEXT     what the whole page is about          │      of the image
           └──────────────────────────────────────────────────────┘             │
           ┌──────────────────────────────────────────────────────┐             ▼
  GOOGLE   │  MACHINE VISION   Google "looks" at the pixels itself  │      → eligible to rank
  ADDS     │  (object, text via OCR, faces, landmarks, products)    │        in Images / packs
  THIS     │  IMAGE INDEX      dedup, quality, where else it appears │        / Lens matches
           └──────────────────────────────────────────────────────┘

Walk the signals top to bottom. The filename is the first and most-ignored: anode-rod-corrosion.jpg tells Google (and a human scanning a URL) something real; IMG_4821.jpg tells it nothing. This is a minor signal — do not oversell it — but it is free, and it is a habit. The alt text is the big one, important enough that it gets all of §11.2. The caption and surrounding text matter because Google reads the image in the context of the words nearest it: an image of a valve on a page titled "Why Your Water Heater Is Leaking," inside a paragraph about the drain valve, is disambiguated by that context. Structured data (the ImageObject, Product, or VideoObject types you will meet in §11.4 and Chapter 18) can attach explicit machine-readable facts. And the page context — the whole topic of the page — colors how Google reads every image on it.

Then there is the row you don't control: Google's own machine vision. Google now analyzes the pixels directly. It can recognize objects, read text inside an image with optical character recognition (OCR), identify landmarks and products, and match a picture against others it has seen. This is the engine behind Google Lens (§11.6), and it is why "the image itself" is now, genuinely, a signal. But — and this is the honest part — you cannot optimize Google's vision model. You can only give it a clear, well-lit, unambiguous subject to look at, and pair it with the words and markup that confirm what it is seeing.

🔎 How Search Sees It When Googlebot processes a page, it does not "watch" your images the way you'd flip through a photo album. It fetches each image URL (which is one reason images can consume crawl resources on huge sites — see Chapter 14), notes the filename and the alt attribute, reads the caption and nearby text, and — as resources allow — runs its vision models over the pixels. Then it files the image in a separate image index, deduplicated against copies elsewhere and scored for quality and relevance. Two consequences fall out of this. First, an image loaded in a way Googlebot can't fetch — hidden behind a script that never runs, or served only after an interaction — may never enter the index at all (§11.3, §11.7). Second, the page the image lives on matters: Google prefers to surface images from pages that are themselves good, relevant, and trustworthy, which means image SEO is never fully separable from ordinary SEO.

A word about EXIF data, because it is the source of a durable myth. EXIF — the Exchangeable Image File format — is metadata a camera embeds inside an image file: the camera model, exposure settings, a timestamp, and sometimes GPS coordinates of where the photo was taken. It is real, and it travels with the file. And it tempts people into one of image SEO's most persistent folk beliefs.

🚫 SEO Myth: "Geotag your images with EXIF GPS data and you'll rank in local search." This one refuses to die, especially in local-SEO circles: the claim that stuffing your business's coordinates (or a keyword) into every image's EXIF metadata boosts local rankings. There is no credible evidence Google uses EXIF GPS as a local ranking factor, and Google's own representatives have repeatedly poured cold water on the idea. Your location signals come from your Google Business Profile, your NAP consistency, and your on-page content (all of Chapter 25) — not from a hidden coordinate in a JPEG. The one embedded-metadata standard Google has publicly committed to reading is IPTC licensing metadata, which powers image credits and the "Licensable" badge — a genuinely useful, documented use we cover in §11.7. So: don't waste an afternoon geotagging your photo library for rankings. Do consider IPTC metadata if you license images. Know the difference, and you've already beaten most of the advice online.

The takeaway for §11.1 is a hierarchy. The signals you control and that reliably matter, in rough order: descriptive alt text, relevant surrounding content and captions, a sensible filename, and (where it applies) structured data. Google's vision model is powerful but not yours to tune. And embedded EXIF keywords or geotags are, for ranking purposes, a myth — with the single, real exception of IPTC licensing data.

11.2 Alt text, done right (accessibility first, image search second)

If there is one image-SEO technique to get right, it is alt text — the alt attribute on an HTML <img> element, a short written description of what the image shows or does. Here is the sentence that should reorder your priorities: alt text exists, first and foremost, for people who cannot see the image. A blind or low-vision person using a screen reader hears the alt text read aloud in place of the picture. A person on a broken connection sees the alt text where the image failed to load. Search relevance is the third beneficiary, not the first — and, happily, the alt text that serves a screen-reader user well is almost exactly the alt text that serves Google Images well. Do it for accessibility, and you get the SEO for free. Do it for SEO at accessibility's expense, and you have both failed a human being and, usually, tripped a spam pattern.

Good alt text describes the image's content and function in its context, concisely and truthfully. A few concrete rules earn their keep:

• Describe what matters here. The same photograph needs different alt text on different pages. On a page about corrosion, "A water-heater anode rod heavily corroded and flaking after six years" is apt; on a page about a technician's visit, "A Rivertown technician holding a spent anode rod beside a customer's water heater" might be right. Context decides.
• Be concise. A sentence or so. Screen readers read every word; a bloated alt is a burden, not a service.
• Don't start with "image of" or "picture of." The screen reader already announces that it is an image; "image of a furnace" becomes "image, image of a furnace." Just describe it.
• Decorative images get empty alt (alt=""). A purely ornamental swoosh or background texture should have an empty alt attribute so the screen reader skips it. Note: empty (alt=""), not missing — a missing alt attribute makes some screen readers read the filename aloud, which is worse.
• If the image is a link or a button, describe the destination or action, not the picture — because for a linked image, the alt text acts as the link's anchor text.

And the anti-pattern, stated plainly, because it is everywhere:

🚫 SEO Myth: "Put your target keyword in every image's alt text to boost rankings." This is keyword stuffing (Chapter 9) wearing a high-visibility vest, and it fails on every axis. To a screen-reader user, alt="furnace repair Cedar Hills emergency furnace repair near me best HVAC" is gibberish read aloud — actively hostile to the person the attribute exists for. To Google, it is a recognizable spam signal in a field Google explicitly warns against overloading. And it does not even work: alt text is a description, and a description crammed with repeated keywords describes nothing. Google's own image guidance says it directly — write descriptive alt text; avoid stuffing keywords. The honest move is to describe the image truthfully, in natural language, and let the relevant words appear because they belong in an accurate description, not because you forced them in.

So where does that leave the SEO value? Real, but bounded and honest.

⚖️ Evidence Check Claim: "Alt text is a ranking factor." Sort it out carefully, because "ranking factor" hides a distinction. — Confirmed by Google (for image understanding): Google's public documentation states that alt text helps Google understand what an image is about and is used to rank images in Google Images. For image search specifically, this is as close to confirmed as image SEO gets. — Weaker for ordinary web ranking: whether good alt text meaningfully lifts a page's position in the normal blue-link results is far less established. Treat any claim that alt text moves web rankings as, at best, a small relevance signal — not a lever. — Confirmed for accessibility (and this is the point): alt text is required for conformance with the Web Content Accessibility Guidelines (WCAG), the international standard, and in many jurisdictions accessibility is a legal obligation. This is the durable reason to do it well — and the reason the SEO benefit is a by-product, not the justification.

Two pieces of good news follow. First, because the accessibility case is airtight, you never have to agonize over whether alt text is "worth it" for SEO — it is worth it regardless, and the image-search relevance rides along. Second, alt text is one of the easiest audits in all of SEO, and you can do it today.

🛠️ Try It on Your Site Open any important page and view its source (or use a free accessibility checker such as the WAVE tool, or your browser's built-in Lighthouse accessibility audit). Look at the alt attributes on your images. How many are missing entirely? How many are alt="IMG_4821" or a stuffed string of keywords? How many actually describe the image in a sentence a blind user would find useful? Pick your five most important pages and fix the alt text on every meaningful image — describe, don't stuff, and give decorative images an empty alt="". You will have improved your accessibility and your Google Images eligibility in the same half hour, with a free tool and no developer.

To make the standard concrete, a comparison — the kind you should be able to make at a glance:

Figure 11.2 — Alt text: the same photo, described four ways (constructed teaching example)

🔄 Check Your Understanding A developer says, "I'll just auto-generate alt text for all 2,000 product images by inserting the product name plus 'buy online' plus the category." Name two distinct problems with this, one for accessibility and one for SEO.

Answer

Accessibility: the generated string ("Blue Trail Jacket buy online Outerwear") is not a description of the image — a screen-reader user learns nothing about what the picture shows (color, style, what's visible), and the "buy online" is meaningless noise read aloud. SEO: it is templated, repetitive, near-keyword-stuffed text that describes nothing and can read as a spam pattern; it also wastes the real opportunity, which is a genuine visual description that helps the image rank in Google Images. Better: generate a truthful base description from real product attributes (color, material, type), and treat "buy online" as page copy, not alt text.

11.3 Image performance: formats, sizing, and lazy loading

Now the technical foundation asserts itself, because images are, on most sites, the single biggest cause of slow pages — and speed and layout stability are part of how Google judges page experience. This section is where image SEO overlaps directly with Core Web Vitals (CWV), Google's set of user-experience metrics. We are going to name the overlap and give you the practical moves, but the full treatment of the metrics themselves belongs to a later chapter, and it would be a disservice to cram it in here.

🔗 Connection Two of the three Core Web Vitals are dominated by images. Largest Contentful Paint (LCP) — how quickly the main content paints — is very often an image (a hero photo, a product shot), so a heavy, slow image is frequently the LCP problem. Cumulative Layout Shift (CLS) — how much the page jumps around as it loads — is very often an unsized image that pushes content down when it finally appears. The metrics, their thresholds, and how to measure them (field data, lab data, PageSpeed Insights) are the whole of Chapter 16 (Core Web Vitals). This section gives you the image-specific fixes; Chapter 16 gives you the instruments and the targets. Mobile-first indexing (Chapter 17) raises the stakes further, because your images have to perform on a phone on a cell network.

Three levers do most of the work.

Format. Choosing the right file format, and a modern one, is the biggest single win. The rough map:

FIGURE 11.3 — Choosing an image format                               [practical guide]
  PHOTOGRAPH (many colors, no transparency) ─▶ WebP or AVIF (modern) ▸ fallback JPEG
  GRAPHIC / SCREENSHOT / TRANSPARENCY needed ─▶ WebP or AVIF        ▸ fallback PNG
  LOGO / ICON / SIMPLE SHAPES (scalable)     ─▶ SVG (vector; tiny, crisp at any size)
  ANIMATION                                   ─▶ video (MP4/WebM), NOT a heavy GIF

  Modern formats (WebP, AVIF) commonly cut file size 25–50%+ versus old JPEG/PNG
  at similar quality — [illustrative range; measure your own].

The practical upshot: serving photographs as WebP or AVIF rather than legacy JPEG, and simple graphics as SVG, can shrink your image payload dramatically at no visible quality cost. As an illustrative example, a 2.4 MB unoptimized hero JPEG re-encoded as a properly sized WebP might land near 180 KB — a more than ten-fold reduction that a phone on a weak connection feels immediately. (Numbers illustrative; your results depend on the image.)

Sizing. Don't ship a 4000-pixel-wide image to display in a 400-pixel box — the browser downloads all four thousand pixels and throws most away. Serve appropriately sized images, and use responsive images (the srcset and sizes attributes, or your CMS's built-in responsive handling) so the browser can pick the right size for the device. Just as important for CLS: always specify each image's dimensions (a width and height, or a CSS aspect-ratio), so the browser reserves the space before the image arrives and the page doesn't lurch. This one habit — dimensioned images — prevents a large share of layout-shift problems.

Lazy loading. Lazy loading defers the loading of off-screen images until the user scrolls near them, so the initial page load isn't burdened with images nobody has looked at yet. It is a real speed win — and the source of a real anxiety.

🔎 How Search Sees It Site owners worry that lazy loading "hides images from Google." Here is the honest mechanism. Google supports standard lazy loading, including the native browser attribute loading="lazy", and when lazy loading is implemented in the standard ways, Googlebot can and does load and index those images. The danger is not lazy loading itself; it is broken lazy loading — an implementation that only loads an image after a real human scroll or click that Googlebot never performs, leaving the image invisible to the crawler. Two safe rules follow. First, use native loading="lazy" or a well-tested method that loads images as they enter the viewport, not one gated behind interaction. Second — and this is the one people get backwards — never lazy-load your LCP image. The big above-the-fold hero is the first thing you want painted; deferring it directly harms Largest Contentful Paint. Lazy-load what's below the fold; eager-load what's at the top.

What image performance can do: make your pages faster and more stable, improving real user experience and removing a page-experience liability that can hold back rankings. What it cannot do: rank a page on speed alone. Core Web Vitals are, in Google's own framing, a modest signal and a tiebreaker among comparably relevant results — not a substitute for being the best answer. A blazing-fast page that answers the wrong question still loses to a slightly slower page that answers the right one. Optimize images because it helps users and clears a hurdle; don't imagine it is the hurdle that matters most. Chapter 16 puts an honest weight on all of this.

11.4 Video SEO: on-site video, VideoObject, and key moments

Video is the visual surface most businesses either ignore or fumble, and it is a shame, because for many queries — how to do something, what a process looks like, whether to trust the person doing it — a short video is the single most persuasive asset a site can own. But "upload a video" and "have a video that search can find and surface" are two very different things, and the gap between them is what video SEO closes.

Start with what Google needs before a video can appear in search at all. Google indexes videos it can find and understand, and its documentation is concrete about the basics: the video should be the meaningful content of a public, crawlable page; it needs a thumbnail (a still preview image); it should use a supported format and a stable URL; and the page around it should describe it in text — a title, a written summary, a transcript where possible. Miss these and the video simply won't be eligible, no matter how good it is. (Google's Video indexing report in Search Console — a Chapter 27 tool — will later tell you whether Google actually indexed each video and why not.)

Then comes the markup that turns an indexed video into a rich result. VideoObject is the Schema.org structured-data type that describes a video to search engines — its name, description, thumbnailUrl, uploadDate, duration, and the URLs where it can be played. Adding it makes your video eligible for video rich results: the thumbnail, duration, and upload date shown right on the SERP. Here is a short, illustrative shape — you are meant to recognize this and hand it to a developer or a plugin, not hand-author it:

{
  "@context": "https://schema.org",
  "@type": "VideoObject",
  "name": "How to Relight a Furnace Pilot Light Safely",
  "description": "A licensed Rivertown technician demonstrates the manufacturer's steps, with safety warnings.",
  "thumbnailUrl": "https://example.com/img/pilot-light-thumb.jpg",
  "uploadDate": "2026-01-15",
  "duration": "PT3M20S",
  "contentUrl": "https://example.com/video/relight-pilot-light.mp4"
}

Read what that does: it hands Google explicit, machine-readable facts it would otherwise have to guess — exactly what the film is, how long it runs, when it was posted, and where the thumbnail lives — which is the whole idea of structured data.

🔗 Connection Structured data is a big enough topic to have its own chapter. VideoObject is defined and used here because it is a video technique, but the general mechanics — what Schema.org and JSON-LD (JavaScript Object Notation for Linked Data) are, how JSON-LD is added to a page, how to test it with the Rich Results Test, and the honest truth that structured data is not a direct ranking factor — all belong to Chapter 18 (Structured Data and Schema Markup). When you implement VideoObject, use Chapter 18's testing loop. Treat this section as the "why and what," Chapter 18 as the "how and prove-it."

Now the feature worth building toward. Video key moments are the timestamped, clickable segments Google can display beneath a video result — little labeled links that jump the viewer straight to "Turn off the gas" or "Relight the pilot" inside a longer video. They are a genuine SERP advantage: they make your result larger and more useful, and they let a searcher land exactly on the part they need. Google can populate key moments two ways: it can automatically detect them (this works especially well when a video has clear chapters, which for a YouTube video means timestamps in the description — see §11.5), or you can mark them up explicitly with structured data (the Clip or SeekToAction properties attached to your VideoObject). Either way, the raw material is the same: a video that is genuinely structured into distinct, nameable steps.

📄 Read the SERP

text FIGURE 11.4 — "A how-to query that wants a video" [constructed teaching example] THE QUERY / PAGE "how to relight a furnace pilot light" — one query, and the results Google returns. WHAT'S THERE Above the text results sits a VIDEO result with a thumbnail, a 3:20 duration badge, a channel name, and — beneath it — four KEY MOMENTS: "Safety first · Find the pilot · Hold the reset · Relight." A "People also ask" box holds related questions. WHAT IT SHOWS Google reads this intent as partly VISUAL and procedural — people want to *watch* the steps, not only read them. The key moments tell you Google values a well-chaptered how-to. This is a video opportunity hiding in plain sight. WHAT IT DOESN'T It does not promise that a new video will get the rich treatment — key moments and video results are Google's call, never guaranteed — and it does not tell you how much traffic the slot earns. The SERP shows demand and format, not a promised outcome. THE MOVE Produce a genuinely useful, safety-first how-to video, structured into named steps, hosted on a crawlable page with a transcript and VideoObject markup — the honest way to become *eligible* for this slot. Then let Google decide. THE LESSON When the SERP shows a video with key moments, the query is telling you it wants to be *shown*, not just told. Match the format, structure the content, and earn eligibility.

Two honesty notes close the section. A hosting decision shadows everything above: a video embedded from YouTube lives on YouTube's terms (and helps your YouTube presence — §11.5), while a self-hosted video on your own crawlable page keeps the on-site SEO benefit but puts the performance burden (§11.3) on you — video is heavy. Many businesses do both: publish on YouTube for reach, and embed on-site with VideoObject for the page. And the limits are real: video rich results and key moments are never guaranteed; production is genuinely expensive in time and skill; and a mediocre video is worse than none, because it burns trust. Video SEO rewards the businesses willing to actually make something worth watching.

11.5 YouTube as a search engine

Here is the reframe that unlocks this section: YouTube is not (only) a video-hosting site. It is a search engine — one with its own query box, its own autocomplete, its own ranking system, and its own colossal audience. People go to YouTube and search: "how to relight a pilot light," "why is my furnace short cycling," "tankless vs tank water heater." If your business has expertise worth demonstrating, those searches are demand you can meet — and most of your competitors aren't even trying.

The claim you will hear attached to this is worth handling with the book's usual care.

⚖️ Evidence Check Claim: "YouTube is the world's second-largest search engine." You will see this stated as flat fact everywhere. — What's solidly true: YouTube is one of the most-visited websites on earth, with a signed-in user base in the billions and an enormous volume of searches; that it is a major search destination, second only to Google web search for many purposes, is not seriously disputed and is directionally right. — What's shakier: the precise ranking "#2 search engine" is a characterization, not a cleanly measured fact. Comparing "searches" across a general web engine, a video platform, a marketplace like Amazon, and an app store is genuinely apples-to-oranges, and the neat ordering papers over real definitional questions. So we will use the strategic truth — YouTube hosts massive, intent-rich search demand you can compete for — and decline to stake anything on the exact rank or an invented percentage. The opportunity is real whether it is "#2" or "#3."

YouTube SEO is the practice of optimizing videos to be found and ranked within YouTube's own search and recommendation systems (and, as a bonus, within Google's video results). The levers are partly familiar and partly particular to YouTube:

• Title. The highest-leverage element: compelling, honest, and built around how people actually search. "How to Relight a Furnace Pilot Light (Safely, Step by Step)" beats "Rivertown HVAC Tips Episode 4."
• Description. The first couple of lines matter most (they show in results and previews). Use the full description to summarize the video, add context, and — critically — place timestamps that create chapters. Those chapters improve the viewing experience and feed Google's key moments (§11.4).
• Thumbnail. The thumbnail — the clickable preview image — is one of YouTube's biggest levers, because it drives click-through rate (CTR), and CTR feeds ranking. A custom, high-contrast, legible thumbnail routinely outperforms an auto-generated frame grab. This is where creators win or lose the click.
• Captions and transcript. Provide accurate captions (not only the auto-generated ones). They serve deaf and hard-of-hearing viewers — the accessibility thread again — and give YouTube and Google clean text to understand the video.
• Engagement signals. This is where YouTube diverges most from web SEO. YouTube's system is driven heavily by watch time and audience retention (do people keep watching?), along with CTR, likes, comments, and whether the video satisfies the viewer. You cannot fake these; you earn them by making a video people actually watch to the end.
• Tags. Once overrated, now minor. Add a few relevant ones and move on; do not spend your energy here.

🛠️ Try It on Your Site Go to YouTube and start typing a question your customers ask — "how to relight" — and watch the autocomplete suggestions populate. That dropdown is YouTube telling you, free, what people actually search for on the platform (the same idea as Google Suggest from Chapter 7, on a different engine). Note the phrasings. Then search one of them and study the top results: their titles, their thumbnails, their lengths, their chapter structure. As with the web SERP, what already ranks is your answer key for what YouTube rewards for that query. You have just done free YouTube keyword research.

Be clear-eyed about the trade-off, though, because YouTube is not a pure gift. You are building on land you don't own: YouTube controls the algorithm, the interface, and the relationship with the viewer, and a policy or ranking change can move your numbers overnight — the same platform-dependence risk this book keeps flagging (theme 6, and Chapter 36's diversification argument). And YouTube's own goal is to keep people on YouTube, not to send them to your site; a video can rack up views while driving few clicks to your booking form. The mitigations are to treat YouTube as one channel among several, to put clear calls-to-action and links in your descriptions and videos, and to also embed the video on your own site (§11.4) so the asset works for you in two places. Made well, a how-to video is the rare content that demonstrates real experience — the "E" in E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) — in a way text struggles to match: it is very hard to fake being a competent technician on camera.

11.6 Visual search and Google Lens: the shift SEOs underrate

Now the frontier, and the part of this chapter most likely to matter more next year than it does this year. Visual search is searching with an image as the query — instead of typing words, you show the search engine a picture and ask, in effect, "what is this, and where can I get it / learn about it / fix it?" Google Lens is Google's tool for exactly that: point your phone camera (or upload an image), and Google identifies what's in the frame — objects, products, plants and animals, landmarks, and any text it can read — and returns results, shopping matches, translations, or answers. On Android, "Circle to Search" bakes the same capability into the whole phone; "multisearch" lets you combine an image and a text refinement ("this jacket, but in green").

Why do SEOs underrate it? Because it is invisible in the tools they stare at. Your rank tracker doesn't show it; your keyword tool doesn't list it; there is no tidy "Lens traffic" line in most reports. But the usage is enormous and growing — Google has publicly described billions of visual searches, and Lens is increasingly a default behavior for a generation that would rather show than type. For some businesses this is already material: a customer standing in front of a mysterious furnace part is far more likely to photograph it than to describe it in words.

🔎 How Search Sees It Lens is the machine-vision engine from §11.1, pointed the other way. Instead of you giving Google an image and Google filing it, the searcher gives Google an image and Google matches it against everything in its image index and Knowledge Graph. It extracts visual features, reads embedded text with OCR, recognizes products against the shopping graph, and identifies entities. The practical consequence for you is subtle but important: to appear in visual-search results, your images need to be in Google's image index in the first place (discoverable, crawlable, not blocked — §11.7), clearly show a single recognizable subject, and — for products — be backed by clean structured data and accurate information. Lens is not matching your keywords; it is matching your pixels and the facts attached to them.

Which sets up the myth this section has to kill.

🚫 SEO Myth: "There's a special way to optimize for Google Lens" (or its cousin, "visual search is a gimmick that doesn't matter"). Both halves are wrong. There is no secret Lens optimization dial — you cannot keyword-stuff a photograph, and there is no lens-rank meta tag. What actually helps is unglamorous and familiar: high-quality, clear, well-lit images with a single obvious subject; accurate structured data (Product and friends, Chapter 18) for anything you sell; images that are actually in the index via clean crawling and sitemaps; and being, generally, a trustworthy source. And "gimmick"? Tell that to any retailer watching camera-first shopping grow. The correct posture is neither chasing a nonexistent trick nor dismissing the surface — it is making your images genuinely machine-legible and your product data genuinely accurate, so that when Google's vision model recognizes what a searcher photographed, your well-described version is the match it surfaces.

📄 Read the SERP

text FIGURE 11.6 — "The camera is the query" [constructed teaching example] THE QUERY / PAGE No typed words. A customer photographs a corroded valve on their water heater and runs Google Lens on the image. WHAT'S THERE Lens identifies it as a "water heater drain valve," offers visually-similar images, a few shopping results for replacement valves, and pages explaining the part and how it fails — including one from a plumbing company that has a clear, well-labeled photo of exactly that valve. WHAT IT SHOWS Google matched the pixels to indexed images and entities. The plumbing company earned a place in a search that contained NO keywords — purely because its image was clear, indexed, and accurately described. WHAT IT DOESN'T It does not tell you the searcher's intent precisely (buy the part? hire a pro?), and it does not report cleanly in analytics — visual-search visibility is real but hard to measure directly. Don't expect a tidy traffic number. THE MOVE Publish clear, single-subject photos of the actual parts and problems you handle, on well-described pages, and (if you sell parts) with Product structured data — so you are *in the set* Lens can match. You can't target the query; you can be eligible for it. THE LESSON Visual search rewards clarity and accurate description, not keywords. In a camera-first world, a genuinely good, well-labeled image is a discovery channel of its own.

The honest limits keep this from becoming hype. Visual-search traffic is hard to isolate and measure, so you are often investing on the strength of the direction rather than a clean number in a dashboard — uncomfortable for a strategist who likes to prove everything. It matters more for some businesses (retail, parts, anything a customer can hold up to a camera) than others. And you cannot control what Lens surfaces any more than you control the web SERP. What you can do is make sure that when visual search comes for your category — and for retail it already has — your images are clear, indexed, described, and backed by accurate data. That is not a trick. It is, once again, just doing the ordinary work well, one surface earlier than your competitors.

11.7 Image and video sitemaps, and licensing metadata

The last piece is discovery — the very first stage of Chapter 1's pipeline, applied to media. You can have the best-described, best-performing images and videos on the web, and it means nothing if Google never finds them. Two families of tools help: sitemaps (to aid discovery) and licensing metadata (to earn credit and the "Licensable" badge).

You met the XML sitemap in Chapter 1 as a discovery aid, and Chapter 14 owns it fully. What matters here is that a sitemap can carry media information, too.

🔗 Connection The XML sitemap — the file that lists your URLs to help Google discover them — is introduced in Chapter 1 and covered properly in Chapter 14 (Technical SEO Fundamentals). Image and video sitemap information is an extension of that same file (or a companion to it), so treat this section as the media-specific layer on top of Chapter 14's foundation. If your sitemap setup is broken, fix that first (Chapter 14); the media extensions only help once the basic sitemap works.

An image sitemap is image information added to your XML sitemap — the image URLs associated with each page — so Google can discover images it might otherwise miss. This is most useful in a specific situation: when images are loaded in ways the crawler struggles with, such as through JavaScript or CSS backgrounds rather than plain <img> tags. For a straightforward WordPress site whose images are ordinary <img> elements, image sitemaps are often unnecessary — Google finds those images by crawling the page. For a JavaScript-heavy gallery or a large catalog, they can genuinely help discovery. Match the tool to the problem; don't add complexity you don't need.

A video sitemap does the same job for video: it tells Google where your videos are and describes them — the play-page URL, the title, the description, the thumbnail, the duration. It is an alternative or a complement to the VideoObject markup from §11.4; you can help Google understand a video through on-page structured data, through a video sitemap, or both. For a site with more than a handful of videos, a video sitemap is a sensible discovery aid.

The crucial honesty, the same as for sitemaps generally: sitemaps aid discovery, not ranking. Listing an image or video in a sitemap helps Google find it; it does nothing to make it rank. A sitemap will not rescue a bad image or an unwatchable video. It removes a discovery bottleneck — no more, and no less.

Then there is the one embedded-metadata standard that genuinely earns its keep, closing the loop on the EXIF myth from §11.1. IPTC photo metadata — a standard from the International Press Telecommunications Council — can embed licensing and credit information inside the image file: the creator, the copyright notice, the credit line, and a URL where someone can license the image. Google reads this metadata and uses it: in Google Images, images with proper licensing metadata can show a "Licensable" badge and a link to license terms, and Google can display the creator and credit. For photographers, stock providers, publishers, and any business whose original images have real value, this is a documented, worthwhile use of embedded metadata — the exact opposite of the geotagging folklore. It doesn't boost rankings; it protects and credits your work, and it surfaces a licensing path to people who want to use your images legitimately.

🔄 Check Your Understanding Two quick ones. (1) You add all your images to an image sitemap and rankings don't move. Is the sitemap broken? (2) A photographer wants credit and a license link to show on their images in Google Images. EXIF geotag, or IPTC metadata?

Answer

(1) No — that's expected. Sitemaps aid discovery, not ranking. If the images were already discoverable (ordinary <img> tags on crawlable pages), the sitemap changes little, and it was never going to lift rankings regardless. (2) IPTC metadata. IPTC licensing/credit fields are what Google reads to show the creator, credit, and the "Licensable" badge with a license link. EXIF GPS geotagging does nothing for this (or for local rankings — §11.1's myth).

📈 The Strategy File

Rivertown Home Services has, at last count, a website full of images that describe nothing and exactly zero videos — which, given that it is a home-services company whose entire value is skilled people doing visible physical work, is a strategic waste. Marisa and Tony Delgado (the second-generation co-owners) have plenty of raw material: technicians on jobs every day, dramatic before-and-afters, and a hundred small how-to moments customers search for. This chapter's increment turns that raw material into a plan. We are planning, not building the schema or the Core Web Vitals fixes yet — those land in Chapters 16 and 18 — and we are not touching the Google Business Profile photos, which are Chapter 25's job.

Part 1 — The image and alt-text plan. A pass over the site's existing images and a standard for new ones:

FIGURE 11.7 — "Rivertown's image + video plan"                       [the Strategy File]
  IMAGE AUDIT FINDING                         THE FIX (per this chapter)
  Service-page photos are stock, generic,     Replace with REAL photos of Rivertown's own
    and shared across all five cities           technicians and jobs (also an E-E-A-T signal, Ch 5);
                                                distinct images per city where honest
  Filenames are IMG_2231.jpg, DSC_0098.jpg    Rename to descriptive: furnace-heat-exchanger-repair.jpg
  Alt text is missing or = the filename       Write true, concise alt text describing each image in
                                                context; empty alt="" for decorative graphics
  Water-heater guide (the page stuck at #11)  Add a clear, well-labeled photo of a corroded anode rod
    has no images at all                        and a leaking drain valve — helps readers, Google
                                                Images, AND Lens (a customer may photograph that valve)
  Hero images are 3–4 MB JPEGs                Convert to WebP, size them correctly, set dimensions;
                                                flag for the Ch 16 CWV pass (don't lazy-load the hero)
  No image discovery aid                      Note for Ch 14: images are ordinary <img> tags, so an
                                                image sitemap is likely unnecessary — confirm later

The alt-text standard we adopt for Rivertown, in one line: describe the image truthfully in a sentence, in the context of its page; never stuff a keyword; give decorative images an empty alt. That standard serves the visually impaired customer first and Google Images second — exactly the priority §11.2 argued for.

Part 2 — The "how to relight a furnace pilot light" video plan. This is Rivertown's first search-driven video, chosen deliberately: it is a real, high-volume, recurring customer question, and it is one Rivertown's licensed technicians can answer with unmistakable authority. The plan:

• Format and safety. A short (~3-minute) how-to, filmed with a licensed Rivertown technician on camera (their name and license noted in the description — real experience, real expertise, Chapter 5's E-E-A-T made visible). Because relighting a gas appliance touches gas safety, the video leads with prominent warnings — if you smell gas, leave and call the gas company; when in doubt, call us — and demonstrates the manufacturer-recommended steps, not improvised shortcuts. Rivertown touches a safety-sensitive topic here and handles it responsibly; the full YMYL-adjacent framing is Chapter 35's.
• Structure into named steps so it can earn key moments (§11.4): Safety first → Find the pilot assembly → Turn the gas control off, then wait → Hold the reset and relight → When to stop and call a pro. Timestamps for each go in the description as chapters.
• Publish in two places. On YouTube (title: "How to Relight a Furnace Pilot Light — Safely, Step by Step"; custom thumbnail; accurate captions; a description with the chapters, safety notes, and a link to Rivertown's booking page) and embedded on the on-site water-heater/furnace guide with VideoObject markup (deferred to Chapter 18's schema work) so the page and the channel both benefit.
• What this does and doesn't settle. It does create a genuine, trust-building asset that meets real search demand on two surfaces and demonstrates experience text can't fake. It does not guarantee a video rich result, key moments, or a #1 YouTube ranking — those are Google's and YouTube's calls (we never promise rankings) — and one video is a start, not a content strategy. It is the first entry in what will become Rivertown's video library.

Filed in the Strategy File: an image/alt-text standard and audit, and a fully specified first how-to video. Both are plans — the schema, the speed fixes, and the profile photos are scheduled for their own chapters, and we have been careful not to get ahead of them.

Conclusion

Search is not, and has not been for years, ten blue links of text. It is images woven through the results, a video slot above the fold, a customer's camera pointed at a broken part, and an entire second search engine called YouTube — surfaces most sites ignore precisely because they require describing what you have rather than gaming what you don't. That is the whole argument of this chapter: visual search is a large, growing, under-contested channel that rewards the same honest work as the rest of SEO. Describe your images truthfully (for the screen-reader user first, for Google Images second). Make them fast and stable, because images are where page speed goes to die. Build video that Google can index and that people actually want to watch, and structure it so key moments become possible. Treat YouTube as the search engine it is. And make your images machine-legible for a camera-first world, without chasing a Lens "trick" that does not exist.

We were honest about the limits throughout, because this field rewards honesty: alt text is a confirmed signal for image search but a weak lever for web ranking; Core Web Vitals is a modest tiebreaker, not a magic bullet, and its full weight is Chapter 16's to assign; "#2 search engine" is a directionally-true characterization, not a measured fact; sitemaps aid discovery but never ranking; and video rich results, key moments, and rankings are all Google's call, never a promise. What is not uncertain is that a genuinely useful, well-described, well-performing image or video is a durable asset — the kind competitors chasing shortcuts will never accidentally build.

Two threads carry forward. The performance work we flagged — heavy images, unsized images, mis-applied lazy loading — is the doorway to Chapter 16 (Core Web Vitals), where speed and stability get measured and weighted honestly. And the VideoObject markup we sketched is a first taste of the structured-data craft that Chapter 18 teaches in full. But next, we stay in Part II and confront a harder truth than "make more things": that publishing more is often the wrong move, and that auditing, updating, merging, and even deleting content — including, someday, some of the images and videos we just planned — is how a site's quality actually rises.

→ Continue to Chapter 12: Content Updating, Pruning, and the Lifecycle of Published Content.

Key Terms

• Image SEO — the practice of helping search engines discover, understand, and surface a site's images, both as results in their own right and as signals for the page they sit on.
• Alt text — the alt attribute on an HTML image: a short written description of what an image shows or does. Its first purpose is accessibility (screen readers, failed-load fallback); it is also Google's primary textual signal for understanding an image and the anchor text for a linked image.
• EXIF (Exchangeable Image File format) — metadata a camera embeds in an image file (camera model, settings, timestamp, sometimes GPS). Not a documented ranking signal; geotagging for local rank is a myth.
• Thumbnail — a still preview image representing a video (or image); a required element for video indexing and, on YouTube, a major driver of click-through rate.
• VideoObject — the Schema.org structured-data type that describes a video (name, description, thumbnail, upload date, duration, play URLs), making it eligible for video rich results. (General schema mechanics: Chapter 18.)
• Video key moments — timestamped, clickable segments Google can show beneath a video result, jumping the viewer to a specific point; populated by Google's automatic detection (e.g., from chapters) or by Clip/SeekToAction markup.
• YouTube SEO — optimizing videos to be found and ranked within YouTube's own search and recommendations (and Google's video results), via titles, descriptions, chapters, thumbnails, captions, and engagement signals such as watch time.
• Visual search — searching with an image as the query instead of text, asking a search engine to identify or find things based on a picture.
• Google Lens — Google's visual-search tool: it identifies objects, text, products, and landmarks in a photograph and returns related results, using machine vision and the image index.
• Image sitemap — image information added to an XML sitemap to help Google discover images (especially ones loaded via JavaScript or CSS); aids discovery, not ranking. (The XML sitemap itself: Chapters 1 and 14.)

Spaced Review

Retrieval practice. Try each before revealing the answer. Some questions revisit Chapters 9 and 10.

1. Name three signals Google uses to understand an image that you control, and one part of image understanding that you cannot tune.
2. Why is "put your target keyword in every image's alt text" wrong on both accessibility grounds and SEO grounds? What is the correct standard instead?
3. Which two Core Web Vitals are most often dominated by images, and what image mistake causes each? (You met the on-page writing craft in Chapter 9 — where does image performance sit relative to being the best answer?)
4. (Chapter 10 revisit.) Chapter 10 taught you to read the dominant format off the SERP. When a results page shows a video with key moments above the text results, what is the query telling you, and what format should you build?
5. (Chapter 9 revisit.) A video's YouTube title is "Rivertown HVAC — Episode 7." Using the title-tag thinking from Chapter 9, rewrite it, and say why your version is better for search.