AI engines cannot see images. ChatGPT, Perplexity, and Claude do not have visual processing active during retrieval crawls — they read text. When an AI engine retrieves your product page to answer "best lightweight marathon shoes with carbon-fibre plates," it reads your alt text, your surrounding paragraph, and your ImageObject schema. If those are empty, generic, or absent, the image contributes nothing to the citation. The product might still be cited from your text description. The image itself is invisible.
Google Lens processes over 12 billion visual searches monthly and grows at 30% annually. Google AI Overviews pull images into answers where visuals improve user understanding. Multimodal queries — users uploading a photo and asking "what is this?" — are growing faster than text queries. The image layer of AI search is expanding while most AEO practitioners focus entirely on text. That gap is an opportunity for brands that act now.
Image AEO is not a separate discipline from text AEO. It is the same entity-signalling, schema-stacking, BLUF-structuring work applied to the visual layer. The principles are identical; only the specific fields differ.
Why Does Alt Text Determine Whether Images Get Cited by AI Engines?
When AI engines encounter an image during retrieval, alt text is the first and most reliable signal for understanding what the image depicts. The HTML alt attribute is readable by every AI crawler that processes web content — GPTBot, ClaudeBot, PerplexityBot, and Googlebot-Extended all parse alt text as primary image context.
AI crawlers do not run image recognition on every image during retrieval. Image recognition is computationally expensive. For most crawls, what the AI engine knows about an image is what the text layer tells it: the alt attribute, the caption text, the filename, and the surrounding 50 to 100 words. A product image with alt text "blue shoes" tells an AI engine almost nothing. The same image with alt text "Blue Brooks Hyperion Elite 5 marathon racing shoes with carbon-fibre plate, 196g, available in UK sizes 6-13" tells the engine exactly what the image shows and whether it matches a buyer's query.
The optimal alt text length is 5 to 15 words — long enough to be descriptive, short enough to be extractable as a standalone signal. AEO-ready alt text includes: the product name or entity being shown, the key distinguishing attribute (colour, material, function), and the relevant context (use case or category). Alt text should never duplicate the surrounding paragraph word-for-word — it should add visual information not present in the text, the same way a good caption does for a human reader.
What Is ImageObject Schema and How Does It Improve AI Citations?
ImageObject schema is a schema.org type that declares machine-readable metadata about an image: what it shows, who created it, when it was published, what page it belongs to, and what the image file URL is. It reduces ambiguity — the core goal of all schema for AI citation purposes.
Pages that combine ImageObject schema with descriptive alt text and relevant surrounding text are more likely to be included in visual AI answers because every signal layer is consistent. The image file, the alt text, the schema declaration, and the page text all say the same thing about the same entity. Consistency is what AI engines need before they are willing to cite an image as a source in an answer where accuracy matters.
{
"@context": "https://schema.org",
"@graph": [
{
"@type": "Article",
"headline": "Image AEO: Alt Text and ImageObject Schema for AI Citations",
"datePublished": "2026-07-03",
"dateModified": "2026-07-03",
"image": { "@id": "https://notioncue.com/blog/image-aeo/#hero-image" }
},
{
"@type": "ImageObject",
"@id": "https://notioncue.com/blog/image-aeo/#hero-image",
"name": "NotioncCue AI citation tracking dashboard showing weekly citation rate by engine",
"description": "Screenshot of the NotioncCue Prompt Tracker weekly citation summary for a B2B SaaS brand, showing citation rate across ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews.",
"contentUrl": "https://notioncue.com/images/notioncue-citation-tracker-dashboard.webp",
"encodingFormat": "image/webp",
"uploadDate": "2026-07-03",
"creator": {
"@type": "Organization",
"@id": "https://notioncue.com/#organization",
"name": "NotioncCue"
},
"license": "https://notioncue.com/image-usage-policy",
"mainEntityOfPage": {
"@id": "https://notioncue.com/blog/image-aeo/"
}
}
]
}Four ImageObject fields carry the most AEO weight. The name field is the image's entity label — it should be the same as or close to the alt text, written as a short declarative description. The description field provides additional context for AI engines that can process longer text — describe what the image actually shows in one to two sentences. The contentUrl must point to a stable, crawlable URL — images behind CDN authentication headers or temporary signed URLs cannot be reliably indexed. The mainEntityOfPage link connects the image to the page entity, telling AI engines this image is authoritative for this page's topic rather than a decorative or incidental element.
What Alt Text Patterns Earn Visual AI Citations?
Alt text earns visual AI citations when it matches the query intent of image-format queries buyers are running. Buyers running image queries on Google Lens or asking visual AI questions follow different patterns from text query buyers.
Product comparison queries. "What does the [product] look like compared to [competitor product]?" Buyers asking this visually want to understand physical differences. Alt text that includes specific visible attributes — dimensions, colour, material, key visual differentiators — earns citations where generic alt text does not. "NotioncCue dashboard citation view, dark mode, five-engine comparison panel" is citable for a screenshot query. "Dashboard screenshot" is not.
How-to and process diagram queries. Users asking "how do I set up X" often receive visual answers in AI Overviews. Diagrams, flowcharts, and annotated screenshots are the highest-value image types for process queries. Alt text for these images should describe the process step the image shows: "Step 3 of AEO prompt tracking setup: adding target prompts to the NotioncCue weekly matrix" is citable. "Setup diagram" is not.
Brand entity queries. When a buyer searches "[brand] screenshot" or "[brand] interface" in Google Images or Google Lens, the images that surface in visual AI answers are those where alt text and ImageObject schema explicitly name the brand and product. This is the visual layer of the entity authority work described in the entity-based AEO guide. Every screenshot, product photo, and interface image on your site should name your brand in the alt text — not because it is keyword stuffing, but because brand name in alt text is accurate entity attribution.
Which Image File Signals Also Affect AI Citation Eligibility?
Three file-level signals that most brands ignore affect whether AI crawlers index images confidently.
Filename. Descriptive filenames with dashes rather than underscores and relevant keywords tell crawlers what the image contains before they parse alt text or schema. notioncue-citation-tracker-dashboard-2026.webp earns indexing confidence. IMG_4392.webp earns no signal. Rename image files on publish — most CMS platforms allow filename editing at upload.
Image format and size. AVIF is the recommended format in 2026 for images where file size matters — 50% smaller than JPEG at equivalent quality. WebP is the fallback. Images that cause LCP delay — loading time for the largest page element — contribute to the core web vitals signal that affects AI crawler prioritisation, as covered in the Core Web Vitals guide. Images that cannot be served because they are too large or behind slow CDN routes are less reliably indexed by AI crawlers than fast-loading images.
Licence and attribution signals. ImageObject schema's license field declares under what terms the image can be used. AI engines that evaluate content for responsible use in AI Overviews and visual answers check whether images carry clear licence and attribution. An image with a Creative Commons licence declaration in its ImageObject schema is more citable in AI visual answers than an unlicensed image that may carry copyright risk.
How Do Visual AI Search and Google Lens Differ From Text Citation?
Visual AI citation earns a different kind of traffic and buyer interaction from text citation. Text citation produces a referral click from an AI engine answer. Visual citation in Google Lens produces direct product discovery — a buyer points a camera at something similar to your product and arrives at your page through visual matching.
For ecommerce specifically, Google Lens results pull from the Google Shopping index. Product images associated with complete Product schema — including offers, aggregateRating, and image fields — appear in Lens results when the image visually matches a buyer's query. The Product schema does not replace ImageObject schema — both should be present on product pages, with the Product schema's image property pointing to the same @id as the standalone ImageObject block.
For editorial content — guides, case studies, research pages — visual citation occurs when AI Overviews include an image from your page alongside a text citation. This happens most reliably when the image is a custom original (not stock photography) that explains something specific to the page topic, and when the ImageObject schema explicitly names what the image shows. Generic stock images with generic alt text never earn AI Overview visual placements. Custom diagrams with descriptive alt text and ImageObject schema earn them consistently for how-to and comparison queries.
How NotioncCue Helps You Track and Fix Image AEO
Image AEO problems are invisible until you check. An AI crawler that cannot read your alt text, cannot reach your image URLs, or cannot find your ImageObject schema silently skips the visual layer — you see no error, no crawl warning, just a lower citation rate that you cannot explain from text-only analysis.
The NotioncCue AI Crawler Audit checks whether images on your key pages are in the server-rendered HTML response (visible to AI crawlers) versus loaded via JavaScript after page render (invisible to most AI crawlers). It specifically flags pages where ImageObject schema is declared but the image URL returns a 403 or redirects to a login screen — one of the most common image AEO failures on enterprise sites with CDN authentication. It also surfaces pages where alt text is empty across informational images, giving you a prioritised list of the highest-traffic pages where image AEO gaps are costing citation potential.
Once the crawler audit surfaces your gaps, the NotioncCue Citation Tracker lets you track whether fixing image alt text and ImageObject schema on your top pages produces measurable citation rate improvement. Because the Citation Tracker monitors what AI engines say about your brand weekly, you can correlate schema and alt text updates with changes in how AI engines describe your product visuals — confirming whether the image layer is now being included in how AI engines understand and recommend your brand.
Start your free NotioncCue trial and run the AI Crawler Audit on your five highest-traffic product or guide pages to see exactly which image signals are missing before you invest time in content improvements that assume the image layer is already working.
Before editing a single alt text attribute, run curl -A "Googlebot" https://yourpage.com | grep -c "alt=" and compare the count to the total number of images on the page. If the numbers do not match, some images are loading via JavaScript and have no alt text in the server HTML response that AI crawlers receive. Fixing the rendering issue is the first step — fixing alt text content on JavaScript-rendered images has no AEO effect until the images are in the initial HTML response.
Frequently Asked Questions About Image AEO and Alt Text for AI Citations
Does Google AI Overviews use images from any page it cites?
Not automatically. AI Overviews select images from pages where the image is closely related to the answer topic, the image is crawlable and indexed in Google Images, and the page has strong overall E-E-A-T signals. Having a high-quality image with descriptive alt text and ImageObject schema on a page that is already cited in AI Overviews increases the probability that the image appears alongside the text citation. The text citation eligibility and the image citation eligibility are independent — a page can be text-cited without any image, or both, depending on whether the image signals pass the visual relevance and indexing checks.
Should every image on a page have ImageObject schema?
No. Apply ImageObject schema only to images that carry substantive informational value: the primary product image, a diagram that explains a process, a chart that shows data, or a screenshot that documents a feature. Decorative images — background patterns, icons, visual dividers — should have empty alt text (alt="") and no ImageObject schema. Applying schema to decorative images adds noise to the schema layer without AEO benefit and creates a potential mismatch between declared schema content and actual page content.
How do you write alt text for charts and data visualisations for AEO?
For charts and data visualisations, alt text should summarise the key finding, not describe the chart type. "Bar chart showing citation rate" tells an AI engine the format. "Pages with HowTo schema earn 1.8x more AI citations than Article-only pages, per NotioncCue 2026 analysis" tells the AI engine the finding — which is what gets cited. Lead with the data point. Include the source attribution in the alt text itself where possible. The surrounding paragraph should describe the chart methodology; the alt text should contain the extractable conclusion.
Does image filename optimisation still matter in 2026?
Yes, but less than alt text and ImageObject schema. Filename is one of the signals AI crawlers use when alt text is absent or minimal — it is a fallback, not a primary signal. Descriptive filenames remain worth doing for the marginal signal they add and for basic image SEO hygiene. The priority order is: fix empty alt text first (highest AEO impact), add ImageObject schema to key images second, then address filenames as part of a systematic image hygiene pass rather than as an urgent AEO priority.