SIA
Syed Irfan Ajmal
Est. 2004  ·  GEO  ·  SEO-PR  ·  Content Marketing
Book a discovery call →
BYLINES & CITATIONS: FORBES · HBR · SEMRUSH · WORLD BANK //// GLOBAL CLIENTS · 4 CONTINENTS //// $160K → $1.2M REVENUE · ORGANIC ONLY ////       
Free Chrome Extension

Can AI see your website?

Many of the sites we check on stage turn out to be partly or fully invisible to ChatGPT, Claude and Perplexity, and their owners had no idea. One click shows the score, the reasons and the fixes. And of the 16 other checkers we reviewed, this is the only one that shows the study behind each rule.

Get early access →See how it scores

By Syed Irfan Ajmal, Fractional CMO · Updated 29 August 2026 · Runs in your browser, nothing leaves it

The extension popup: an AI visibility score of 100 for syedirfanajmal.com, split 55 site-wide and 45 this page, a grade C for the page, and a Show on page button

Sample.The extension run on this site’s own Authority Flywheel page, 28 Aug 2026: visibility 100 (site-wide files 55 of 55, that page 45 of 45), answer readiness grade C. Yours will differ.

Every line shows the receipt

The sentence, the number and the study, not a mystery score#

Real lines from real checks. Each one names the place on the page, gives the measurement, and cites the rule, so a writer knows what to change and a client can verify the reasoning.

Answer-first check on an article
Under "Bring this to your stage or team": first sentence runs 27 words ("I give talks and workshops on earned media and the Authority Flywheel, including..."). Fix: open each section with one plain sentence of 6 to 20 words.
Instead of: answer density: 72
Sentence-length check on a news story
Longest: 44 words: "It's resulted in a public that's a little more on edge and panic about what's going on...". No cited sentence in 42,971 was over 17.
Instead of: readability: 58
Honesty check on a page with hidden FAQ markup
FAQPage schema lists 2 questions but 2 are not in the visible text (e.g. "How much does the widget cost?"). Google's policy: markup must match what people see.
Instead of: FAQ schema: present (+10)
Honesty check on a page with a bumped date
The date robots see (2026-08-20) is newer than the date people see (2023-02-01). Keep them honest.
Instead of: freshness: good
Two results, one click

A visibility score, split site and page. A grade for the page.#

Some visibility checks belong to the whole site: robots.txt, llms.txt and the sitemap are the same on every page, and they carry 55 of the 100 points. The other 45 are measured on the page you have open, because JavaScript rendering and metadata differ page by page. The popup shows both numbers, so two pages of one site are never mistaken for two verdicts on the site. Whether AI would quote you is a third question, about one page only, so it gets its own grade instead of being folded in.

AI VISIBILITY · 0 TO 10010055 site-wide: robots.txt · llms.txt · sitemap45 this page: JavaScript · metadataShown as two numbers in the popupANSWER READINESS · A TO ECanswer shape · substance · trustten checks, each with evidenceThis page only, changes page by pageone click
Crawler access
40 of 100 · site-wide
Does robots.txt let the 16 AI crawlers in, or tell them to stay out? Search and user-triggered bots weigh more than training bots, because blocking those is what removes you from answers.
Why this weight: OpenAI and Anthropic each document which of their bots feed search answers and which only train models; the split is theirs, not ours. 1617
JavaScript visibility
30 of 100 · this page
AI crawlers read the raw HTML, before scripts run. We compare the raw copy of the open page with the finished page and report what share of its words is invisible ink. Measured per page: a script-heavy product page and a static article on the same site score differently.
Why this weight: Google's own guidance: if a page cannot be indexed with a snippet, it cannot appear in AI features. 4
llms.txt
5 of 100 · site-wide
A menu card for AI and agent readers. Checked for presence and format.
Why this weight: Deliberately only 5 points. Across 137,000 domains, 97% of llms.txt files were never requested by any bot, and no study has found a citation correlation. Kept because OpenAI and Anthropic recommend it for agent workflows. 12
Metadata and sitemap
25 of 100 · 15 this page, 10 site-wide
Title, description, canonical, Open Graph and structured data are read from the open page (15 points). The sitemap is one file for the whole site (10 points).
Why this weight: Indexability basics. Structured data is counted here as a filing aid, not as an AI-citation lever (see what we refuse to score). 4
The page grade, A to E

Ten checks. Each one has a study behind it.#

AI answers are assembled from pages that already look like answers. In 2025 and 2026, several teams measured which page features actually correlate with being cited, across tens of millions of citations. These ten checks are the ones the data supports, weighted by how strong the evidence is. Content pages only: a homepage or pricing page shows “n/a” rather than a bad grade.

A80+ of 100
B65 of 100
C50 of 100
D35 of 100
Eunder 35 of 100
CheckPointsWhat it looks forWhat the data found
A short, direct answer near the top of each section15One plain sentence of 6 to 20 words among the first two of each section, not a filler opener like 'In this article we will explore'.44% of ChatGPT citations come from the first 30% of a page. The exact sentences AI quotes average 9.8 words and never exceed 17. 67
Sections of 80 to 250 words10Word count between one heading and the next. Stubs under 50 words and walls over 250 both lose points.Sections of 120 to 180 words average 4.6 citations; sections under 50 words average 2.7. 25
Sentences of 20 words or fewer8Share of sentences at or under 20 words. The report quotes your longest sentence.No cited sentence in 42,971 citations was longer than 17 words; 92% were under 20. 7
Lists, tables and definition sentences12Lists, tables and 'X is a...' sentences, about one per 400 words.Pages with lists, tables or headings were matched to AI answers 91% of the time; prose-only pages 39%. 711
Specific numbers and statistics12Data points inside real sentences, per 1,000 words. Label numbers in graphics do not count.Pages with 19 or more data points average 5.4 citations; pages with few average 2.8. 2
Quotes attributed to a named person8Quoted spans of four words or more with a name and a cue such as 'said' or a job title nearby.Pages with expert quotes average 4.1 citations; pages without, 2.4. 2
A visible, recent updated date10The newest date on the page or in its schema; full marks within 3 months. A date that lives only in the schema, invisible to readers, earns at most 6 of 10. A schema date newer than the visible date, or one that equals the scan date, is flagged.Content updated within three months averages 6.0 citations vs 3.6. AI-cited content is 25.7% fresher than organic results, except on Google AI Overviews, where freshness is neutral. 29
A named author readers can see12A visible byline, plus a credential line and a profile link. An author only in schema gets a quarter of the points.Visible expertise and authorship signals were 30.6% more common on cited pages than on uncited ones, the second-largest text difference in the study. 8
Named people, companies, products and places8Share of capitalised names among all words.Cited passages run about 20% proper nouns; ordinary text runs 5 to 8%. 6
No sales language5Share of sentences with promotional phrases ('game-changing', 'unlock', 'cutting-edge') or exclamation marks.Promotional tone was the only text parameter negatively associated with being cited: 26% less common on cited pages. 8
Total100Answer shape carries 45 points, substance 30, trust 25.
A short, direct answer near the top of each section
15 pts
One plain sentence of 6 to 20 words among the first two of each section, not a filler opener like 'In this article we will explore'.
Data: 44% of ChatGPT citations come from the first 30% of a page. The exact sentences AI quotes average 9.8 words and never exceed 17. 67
Sections of 80 to 250 words
10 pts
Word count between one heading and the next. Stubs under 50 words and walls over 250 both lose points.
Data: Sections of 120 to 180 words average 4.6 citations; sections under 50 words average 2.7. 25
Sentences of 20 words or fewer
8 pts
Share of sentences at or under 20 words. The report quotes your longest sentence.
Data: No cited sentence in 42,971 citations was longer than 17 words; 92% were under 20. 7
Lists, tables and definition sentences
12 pts
Lists, tables and 'X is a...' sentences, about one per 400 words.
Data: Pages with lists, tables or headings were matched to AI answers 91% of the time; prose-only pages 39%. 711
Specific numbers and statistics
12 pts
Data points inside real sentences, per 1,000 words. Label numbers in graphics do not count.
Data: Pages with 19 or more data points average 5.4 citations; pages with few average 2.8. 2
Quotes attributed to a named person
8 pts
Quoted spans of four words or more with a name and a cue such as 'said' or a job title nearby.
Data: Pages with expert quotes average 4.1 citations; pages without, 2.4. 2
A visible, recent updated date
10 pts
The newest date on the page or in its schema; full marks within 3 months. A date that lives only in the schema, invisible to readers, earns at most 6 of 10. A schema date newer than the visible date, or one that equals the scan date, is flagged.
Data: Content updated within three months averages 6.0 citations vs 3.6. AI-cited content is 25.7% fresher than organic results, except on Google AI Overviews, where freshness is neutral. 29
A named author readers can see
12 pts
A visible byline, plus a credential line and a profile link. An author only in schema gets a quarter of the points.
Data: Visible expertise and authorship signals were 30.6% more common on cited pages than on uncited ones, the second-largest text difference in the study. 8
Named people, companies, products and places
8 pts
Share of capitalised names among all words.
Data: Cited passages run about 20% proper nouns; ordinary text runs 5 to 8%. 6
No sales language
5 pts
Share of sentences with promotional phrases ('game-changing', 'unlock', 'cutting-edge') or exclamation marks.
Data: Promotional tone was the only text parameter negatively associated with being cited: 26% less common on cited pages. 8
Total 100: answer shape 45, substance 30, trust 25.

The biggest factor of all, how often other websites mention you, cannot be read from one page: across 75,000 brands it correlates with AI Overview presence at 0.66, three times more than backlinks 10. That is the question the Earned Media OS exists to answer.

Check my site: get early access →
Equally important

What it refuses to score, and why#

Most AI-visibility tools still reward things the data says do not matter. This one lists them, gives them zero points, and says so on every result.

Schema markup (FAQPage, HowTo, Article) as a citation lever
In the closest thing to a controlled test, 1,885 pages that added JSON-LD were compared with 4,000 matched controls for seven months: AI Overview citations fell 4.6% and ChatGPT and AI Mode did not move. Google removed FAQ rich results in May 2026 and says no special structured data is needed for AI features, so we list your schema, warn when it does not match the visible page, and give it no points. 123415
Question-shaped headings
Pages with question headings averaged 3.4 ChatGPT citations vs 4.3 for plain headings. What gets quoted is the direct answer under the heading, whatever its grammar, so we count them and say 'fine either way'. 25
Word count
Across 174,000 cited pages, length correlates with AI Overview citation at 0.04, and more than half of cited pages are under 1,000 words. We show the count and score nothing. 13
Links to other websites
Whether a page links to domains of trust 70 or 100, the measured influence on being cited was almost zero. Nice for readers; not a lever. 2
Keyword-rich URLs and titles
Across 1.03M URLs, path depth, length and 'how-to' or 'best' patterns showed no relationship with citations, and heavily keyword-matched titles were cited less. Not checked at all. 14
Reading grade level
Citations are bimodal: very easy and very dense pages are both cited more than the middle. Sentence length is the usable signal, so that is what we score. 7

This is also how the tool itself changed. An external review pointed at the llms.txt data and its weight dropped from 10 points to 5. The answer-readiness spec started with a 30-point schema bucket and a question-heading check; the research pass killed both before a line was built.

How it compares

Against Adobe’s checker and the typical AEO extension#

Adobe’s AI Content Visibility Checker does one thing well: it shows how much of a page survives with JavaScript off. The small AEO and GEO extensions on the Chrome Web Store mostly hand back a number. Check by check:

CheckAI Visibility CheckerAdobeTypical AEO extension
Which AI crawlers robots.txt blocks, with the exact lineYes, 16 botsNoPartly
Text hidden behind JavaScriptYes, 30 of 100 pointsYes, its whole productSome
A grade for the page's shape as an answer, with the sentence behind each verdictYes, ten checksNoA number only
The study and figure behind each rule, on screenYesNoNo
Says what it refuses to score, and whyYesNoNo, most still reward schema
Warns when FAQ schema or a schema date does not match the visible pageYesNoNo
Knows a homepage is not an article (shows n/a instead of a bad grade)YesNoNo
Paints facts, filler and long sentences on the live pageYes, per sentenceHighlights hidden blocksOne tool, per sentence
Corrected robots.txt, starter llms.txt, client report, share card, compare viewYesNoSchema generators
Tracks whether AI actually mentions your brand
The biggest factor in every study, and unmeasurable from one page.
No, on purpose: that is EMOSIn its paid parent productSome, with your own API key
Which AI crawlers robots.txt blocks, with the exact line
This toolYes, 16 botsAdobeNoTypicalPartly
Text hidden behind JavaScript
This toolYes, 30 of 100 pointsAdobeYes, its whole productTypicalSome
A grade for the page's shape as an answer, with the sentence behind each verdict
This toolYes, ten checksAdobeNoTypicalA number only
The study and figure behind each rule, on screen
This toolYesAdobeNoTypicalNo
Says what it refuses to score, and why
This toolYesAdobeNoTypicalNo, most still reward schema
Warns when FAQ schema or a schema date does not match the visible page
This toolYesAdobeNoTypicalNo
Knows a homepage is not an article (shows n/a instead of a bad grade)
This toolYesAdobeNoTypicalNo
Paints facts, filler and long sentences on the live page
This toolYes, per sentenceAdobeHighlights hidden blocksTypicalOne tool, per sentence
Corrected robots.txt, starter llms.txt, client report, share card, compare view
This toolYesAdobeNoTypicalSchema generators
Tracks whether AI actually mentions your brand
The biggest factor in every study, and unmeasurable from one page.
This toolNo, on purpose: that is EMOSAdobeIn its paid parent productTypicalSome, with your own API key

Who is in the “typical” column: SAGE (828 users), RankSEO (688), SellOnLLM (619) and Zicy (456), per their Chrome Web Store listings on 28 August 2026; Adobe's AI Content Visibility Checker listed 10,000+ users the same day. Columns describe public listings and documentation, not hands-on tests.

Fair is fair: Adobe’s rendering check is polished and its paid parent product tracks real AI answers at scale, which this extension deliberately does not. If you need answer tracking, that is the EMOS platform, not a browser button.

Three ways to fix your score
01

Fix it yourself, free

The extension hands you a corrected robots.txt and a starter llms.txt on the spot, and the downloadable report explains every finding in plain words. Most sites can be fixed in an afternoon.

Browse the free playbooks →
02

Learn the system

Being visible is step one. Being cited by AI when buyers ask is the game. EMOS Academy teaches founders the Earned Media OS: how coverage, authority and AI visibility compound.

See EMOS Academy →
03

Have it done for you

A fractional CMO engagement puts 22+ years of earned-media practice on your team: the fixes, the strategy and the coverage that makes AI quote you.

Book a discovery call →
Sources

The studies behind every number on this page#

All of these are correlation studies unless noted; only the Ahrefs schema test approaches a controlled design. Sample sizes are given so you can weigh them yourself.

Show the 18 sources
  1. [1]Ahrefs, 2026. Schema markup and AI citations: 1,885 pages that added JSON-LD vs 4,000 matched controls, Aug 2025 to Mar 2026. ahrefs.com/blog/schema-ai-citations/
  2. [2]SE Ranking, 2026. How to optimize for ChatGPT: 129,000 domains, 216,524 pages, 20 niches. seranking.com/blog/how-to-optimize-for-chatgpt/
  3. [3]Search Engine Journal, 2026. Google drops FAQ rich results from Search (7 May 2026). searchenginejournal.com/google-drops-faq-rich-results-fr…
  4. [4]Google Search Central. AI features and your website: no special structured data or AI files required. developers.google.com/search/docs/appearance/ai-features
  5. [5]SE Ranking, 2026. How to optimize for AI Mode: 2.33M pages, 295K domains, 500K prompts. seranking.com/blog/how-to-optimize-for-ai-mode/
  6. [6]Kevin Indig via Search Engine Land, 2025. ChatGPT citation study: 18,012 sentence-verified citations from 3M responses. searchengineland.com/chatgpt-citations-content-study-469…
  7. [7]Shashko, 2026. Grounding-citation analysis: 42,971 citations on six AI platforms, 11,672 exact cited sentences. hackmd.io/@A09fyOMpSD2VYIJodmXHqQ/r1eJyqthdbe
  8. [8]Semrush, 2025. Content optimization for AI search: 304,805 cited URLs vs 921,614 ranking URLs, 13 text parameters. semrush.com/blog/content-optimization-ai-search-study
  9. [9]Ahrefs, 2026. Do AI assistants prefer fresh content: 17M citations across 7 platforms. ahrefs.com/blog/do-ai-assistants-prefer-to-cite-fresh-co…
  10. [10]Ahrefs, 2025. Brand web mentions vs AI Overview visibility: 75,000 brands. ahrefs.com/blog/ai-overview-brand-correlation/
  11. [11]Surfer, 2025. AI Overviews study: 405,576 AI Overviews. surferseo.com/blog/ai-overviews-study/
  12. [12]Ahrefs, 2026. llms.txt study: 137,000 domains with server logs. ahrefs.com/blog/llmstxt-study/
  13. [13]Ahrefs, 2025. Short vs long content in AI Overviews: 174K cited pages. ahrefs.com/blog/short-vs-long-content-in-ai-overviews/
  14. [14]Otterly, 2026. URL structure and AI citations: 1.03M URLs. otterly.ai/blog/url-ai-citations-study/
  15. [15]Google Search Central. Structured data general policies: markup must match visible content. developers.google.com/search/docs/appearance/structured-…
  16. [16]OpenAI. Overview of OpenAI crawlers: OAI-SearchBot, ChatGPT-User, GPTBot. developers.openai.com/api/docs/bots
  17. [17]Search Engine Journal, 2026. Anthropic's Claude bots: ClaudeBot, Claude-User, Claude-SearchBot. searchenginejournal.com/anthropics-claude-bots-make-robo…
  18. [18]Aggarwal et al., KDD 2024. GEO: Generative Engine Optimization (lab benchmark; later replication found most methods ineffective). arxiv.org/pdf/2311.09735

↑ Back to the checks

Get the extension

The AI Visibility Checker is heading to the Chrome Web Store. Until the listing is live, catch it in action at Irfan's talks and workshops, or ask for early access.

Get early access →Privacy policy