AI Visibility Scorecard
goodatlooking.com · Media & Entertainment
Borderline visible. AI bots can crawl Good at Looking, but the structured-data signals are thin - you are at real risk of being skipped when buyers ask ChatGPT, Claude, or Perplexity for a recommendation.
AI engines read this profile 4 times
Claude · Meta AI · ShapBot
#1,491,573 of 1,966,415 in Media & Entertainment for AI visibility
<1
4
AEO Visibility
iNear-invisible · 3.6/10
5
Muse Index Score
iAI agent readiness
Basic · 5/100
0
AI Adoption
iNone detected · 0/100
from our crawl and measurement
Good at Looking is a business whose own site puts it this way: "It asks, ‘How can textiles unpack, question, unspool, unravel and therefore reimagine the world around us?’ Two works in the Barbican’s exhibition are particularly powerful expressions of." When AI engines look at it, they find little to work with. It scores 3.6 out of 10.
On its own pages the site does well on a clear heading structure. It is passable on a few credibility markers, basic crawler access and partial sitemap and link coverage. What holds it back is a clearly stated business identity and structured data describing the business.
Off the page, other sites link to it. Beyond that, the domain does not yet read as established, AI engines do not yet recognise it as a distinct business, there is no press coverage to draw on, nobody is discussing it where AI engines look and it does not come up on Reddit. None of that can be fixed on the page: it has to be earned.
AI crawlers have visited once in our tracking, including ClaudeBot (Anthropic).
Strong · Good · Fair · Weak
Structured Data
Weak
Organization / LocalBusiness JSON-LD that AI can read.
Content Structure
Good
Clear headings and answer-style content.
Entity Clarity
Weak
How clearly your brand identity reads to AI.
E-E-A-T Signals
Experience, Expertise, Authority, Trust
Fair
Experience, Expertise, Authority, Trust markers.
Technical AEO
Fair
robots.txt, llms.txt, and AI-bot crawl access.
AI Discoverability
Good
Sitemaps and entity links AI can follow.
How the web signals your brand to AI
Backlinks
Fair
Inbound links from other sites.
Domain Authority
Weak
Little domain authority yet.
Reference Presence
Weak
Not in AI knowledge graphs yet.
News & Press
Weak
No press coverage found yet.
Community
Weak
No community discussion yet.
Social Mentions
Weak
No social discussion found yet.
Your AEO score measures whether AI search engines - ChatGPT, Claude, Perplexity, Gemini - can actually read your site and cite it in answers. Roughly two-thirds of sites are invisible to them. At 3.6/10, Good at Looking is crawlable but under-signaled - fixable, and the signals above are where to start.
Can design help us have difficult conversations?
23/12/2017 2 Comments The year of 2017 will be bookended by women speaking out about sexual harassment. And these bookends are both captured visually in two designs created by women: the pink pussyhat and the #metoo hashtag. Both were created as visual symbols of, and means of communicating, women's anger and outrage. They've been hugely influential, spearheading vital and timely campaigns.
Is this your brand?
The exact fixes for Good at Looking
Which AI engines already crawl you
Free AI bot monitoring: see every AI crawler that visits you
Already have an account? Sign in
The search landscape has fundamentally shifted. While Google still dominates, millions of users now ask questions to ChatGPT, Gemini, and Claude instead of typing into a search bar.
Continue reading in your free Engagemii portalFree signup unlocks the full article plus your personalized AEO fix list for Good at Looking.
Scored by Engagemii on August 1, 2026. Methodology: engagemii.com/aeo/methodology
Source URL: https://engagemii.com/aeo/brands/goodatlooking
Cite this score: Engagemii (2026). "AEO Score for Good at Looking." Retrieved from https://engagemii.com/aeo/brands/goodatlooking
Licensed under CC BY 4.0. You may reuse this data with attribution: a visible link to engagemii.com.
Powered by Engagemii - The Answer Engine Optimization (AEO) Platform