AI visibility checker: see what ChatGPT, Claude and Gemini can read.
Lantad shows what AI can actually read on your pages, and what to fix to rank above competitors.
Lantad is an AI visibility checker and generative engine optimization (GEO) tool. It checks a page the way an AI crawler reads it.
Evidence, not vibes.
One URL produces three things: the run itself, a graded score with the sub-scores that built it, and a fix list in the order worth working. The tiles below are a worked example, drawn from a sample scan.
Run it from the browser
Paste a URL and watch the run. Lantad reads the page the way an AI crawler does, then grades what came back.
Example output. Your scan reports your own figures.
Example figures from a sample scan, not live data. The four weights are the real formula, published in full on the methodology page.
Ranked by impact
Every finding lands as an action with the evidence that caused it. The catalogue holds 24 named defects, and a report only ever raises the ones it measured.
Example defects. Your report ranks the ones found on your page.
Every finding in a real report links to the thing that caused it: the exact robots.txt line, the exact words a crawler could not see, the exact response your server returned. Anything the scan could not measure is reported as not measured, with the reason, rather than folded into the grade as a guess.
AI answers are the new front page.
More people ask ChatGPT, Claude, Gemini, and Perplexity instead of scrolling a page of links. Those answers are built from what AI crawlers could read on your site, not from what your visitors see. If a crawler receives a blank page, your brand is missing from the answer, and nothing in your analytics will tell you so.
One page, two readers
Same URL, same server, two different documents. Prose Parity is the size of the gap between them, and the score below is what that gap costs you.
You cannot be recommended if you cannot be read.
About 15 percent of daily views to English Wikipedia
articles went away where Google AI Overviews rolled out.
Wikipedia is one of the sources those answers cite most.
Source: Khosravi and Yoganarasimhan, University of Washington, Impact of AI Search Summaries on Website Traffic, arXiv 2602.18455, posted 5 February 2026 and revised 12 May 2026. Read 7 August 2026. Lantad did not run this study and holds no traffic data for any site.
About 85% of daily views stayed About 15% went away
The split is the cited study's estimate for daily views to English Wikipedia articles where Google AI Overviews rolled out. It is not a Lantad measurement, and the full source is directly below.
Answers, not links
AI assistants summarise instead of linking. If your content is not readable, you are not cited, and the click never happens.
JavaScript is the blind spot
Most AI crawlers do not run JavaScript. Anything that appears only after your app hydrates simply is not there for them.
A new surface, new rules
Being readable to AI is becoming as important as being indexable by Google was. Generative engine optimization needs its own checks.
Prose Parity
Prose Parity is the share of your rendered, human-visible content that ships in the initial HTML an AI crawler reads. One hundred percent means a crawler sees everything your visitors see. A low score means content is arriving via JavaScript that most AI crawlers never run.
No other AI visibility tool reports this. Rank trackers tell you where you placed. Prose Parity tells you whether you were legible enough to place at all.
One page, line by line
A worked example: 8 lines of a product page, in the order a visitor reads them. Every line appears once, and its own row says what becomes of it in the initial HTML. The bar beside each line is that line's word count, because parity is measured by word rather than by block, so the working behind the score sits on the face of the figure instead of underneath it.
- Acme Analytics In the HTML 2 words
- Real-time dashboards for product teams. In the HTML 5 words
- Acme turns raw product events into dashboards your whole team can read, without waiting on a data engineer. In the HTML 18 words
- Trusted by 4,000 engineering teams. Social proof widget, mounted after load Never arrives 5 words
- Pricing starts at $29 per seat, per month. Pricing fetched inside a client component Never arrives 8 words
- Connects to Postgres, Snowflake, and BigQuery. In the HTML 6 words
- What is Acme? An analytics layer that sits on top of your existing warehouse. FAQ accordion, panel text added on click Never arrives 14 words
- Book a demo In the HTML 3 words
34 words reach the crawler 27 words never arrive
56% Prose Parity, 34 of 61 words
The score is arithmetic, not a judgement: 34 of the 61 words in this example reach the crawler, so the score is 56%.
Your report shows your own two corpora and the words that separate them, so any figure on it can be checked line by line.
The measure is one-directional on purpose: the question is how much of what a visitor sees a crawler receives, not how closely the two documents resemble each other.
How to read your score
Crawlers receive effectively the whole page. Clear the remainder when convenient and move on to structure and citations.
Your headline survives and your proof does not. The most common band, and the most expensive, because the page looks fine to everyone who checks it in a browser. Nobody reports it, because nobody can see it.
An assistant answering a question about your category cannot use you as a source. You are absent from the answer, not ranked low in it.
What gets counted
Prose Parity counts words, and it counts them equally. The first three cards say what the extractor treats as main content and what it leaves out. The fourth is the boundary of the metric: structured data is scored on its own, and the composite weights quoted there are read from the same configuration the scoring code reads, so this section cannot drift away from the scanner.
The same boundary on both passes.
A menu that survives while the article does not cannot lift the score, because the menu was never in the corpus to begin with.
Body copy
Every paragraph and list item a visitor can read, measured by word count rather than by block count, so one missing paragraph costs more than one missing label.
Headings
h1 through h6 count as main content wherever they sit. Headings are how an assistant works out what a page is about, so a heading that mounts late costs more than its word count suggests.
Navigation and chrome
Text inside nav, header, footer and aside is boilerplate and stays out of the parity corpus, so a big menu cannot inflate a thin page into a healthy score.
Structured data
JSON-LD present in the source is its own sub-score, not part of parity. Parity is 50% of the composite, crawler access 25%, structure 15%, structured data 10%.
Where the loss usually comes from
Six patterns account for most of the parity Lantad measures as missing. Each fix links to the guide for that stack.
-
Next.js
Content fetched inside a client component
Move the fetch to a server component, or pre-render the route.
-
React, Vue
A single-page-app shell with no server rendering
Add server rendering, or generate static HTML at build time.
-
Shopify
Review, FAQ and recommendation apps injected after load
Render the app content into the Liquid template, or mirror it in the source.
-
WordPress
Page builders that lazy-render everything below the fold
Disable lazy render for text blocks. Images can stay lazy, copy cannot.
-
Any stack
Tabs and accordions that mount their panel text on click
Ship the panel text in the HTML and hide it with CSS, or use a native details element.
-
Any stack
Consent walls and bot challenges served ahead of the content
Let named AI crawler user agents through at the edge, or serve them the same HTML.
Fetch. Render.
Compare. Grade.
Two passes over the same URL, then the difference between them becomes your score and your fix list. Nothing in the grade is a black box.
-
Fetch as a crawler
LantadBot requests your page with no JavaScript execution, the way GPTBot, ClaudeBot, and PerplexityBot do, and identifies itself honestly while doing it.
-
Render in a browser
The same URL loads in a real browser, scrolled and settled, so there is a full hydrated version of the page to measure the first pass against.
-
Compare the two
Prose Parity, headings, links, and structured data are diffed. Anything present in one pass and missing from the other is flagged with the words that went missing.
-
Grade and rank
You get an AI Visibility Score, the robots.txt rule for each of the 15 named crawlers, a live probe result for the 12 that publish a user agent, and fixes ordered by impact.
- Prose Parity 50%
- Crawler access 25%
- Structure 15%
- Structured data 10%
The composite is published, not implied: score = 50% Prose Parity plus 25% crawler access plus 15% structure plus 10% structured data.
Those four numbers are read from the same configuration the scoring code reads, and the methodology page shows the working behind each one.
Built for the AI layer,
not bolted onto SEO.
Classic SEO tools check Google, which renders JavaScript. Lantad checks the layer underneath it, where most crawlers do not.
Two passes, one diff
A non-rendering crawler fetch and a real browser render of the same URL, diffed against each other. The gap between them is the measurement.
Named crawlers, live
15 AI crawlers resolved against your robots.txt, then 12 of them probed with their own published user agent, so a CDN rule cannot hide behind a permissive file.
Fixes, not a dashboard
Every finding lands as a ranked action with the exact tag, file or rule to change, written for the person who will make the change.
Honest by default
The baseline fetch always identifies as LantadBot. A page we cannot grade fairly says so instead of inventing a number, and site owners can opt out.
Every token read on every scan
- GPTBot
- OAI-SearchBot
- ChatGPT-User
- ClaudeBot
- Claude-SearchBot
- Claude-User
- anthropic-ai
- PerplexityBot
- Perplexity-User
- Google-Extended
- Applebot-Extended
- Meta-ExternalAgent
- Bytespider
- CCBot
- Amazonbot
Most AI visibility platforms report where you rank in AI answers. That is useful once you are in them. Lantad starts one step earlier and answers whether your words reach the crawler at all, because a page that cannot be read cannot be ranked.
Anyone whose buyers
ask an AI first.
One page or a hundred, in-house or on the agency side, the question is the same: can an AI crawler read the thing you want it to repeat. Each card below lands on the page that answers it for that job.
SEO and GEO teams
You rank on Google and still get no AI citations.
- Prove the gap with a graded score
- Take ranked fixes to engineering
- Read the formula behind every point
Engineering teams
You own the render path and need to know what ships in the HTML.
- Prose Parity per template
- Scans from code over a REST API
- An MCP server, so an assistant can run the check
Founders and marketers
You need to know whether AI recommends you or a competitor.
- What a named model says about your brand
- A shareable report link, no login to read
- No setup, no sales call
Agencies
You need a repeatable audit that clients understand.
- Up to 8 pages in one run
- Ranked, defensible findings per page
- Re-verify after every deploy
Choose how to get visible
Either route runs the same loop. The difference is who does the middle step.
- Paste a URLNo signup, no email, no card
- Read the scoreUnder a minute, p95 target
- Work the fix listRanked by impact, with the evidence
- Scan againVerify, then repeat on the next deploy
Scan on your own
Run the checker, work the ranked fix list, then verify again after every deploy. The scan itself stays free.
Get it engineered
A Deep Audit of the whole site, a Fix Sprint on your money pages, then a retainer that keeps watching as the site changes.
Lantad is a free AI visibility checker. Lantad fetches a page as a non-rendering AI crawler, renders the same URL in a browser, compares the two, and grades the gap. Every scan returns an AI Visibility Score, a Prose Parity percentage, the robots.txt rule for 15 named AI crawlers, and a ranked fix list.
Most AI crawlers do not run JavaScript. A page can look complete on screen, rank on Google, and still hand ChatGPT an empty shell.
Questions, answered
What is an AI visibility checker?
An AI visibility checker shows what an AI crawler can actually take from a page, which is not what a browser shows you. Most AI crawlers do not run JavaScript, so a site can look complete on screen and hand ChatGPT or Perplexity an almost empty shell. Lantad is an AI visibility checker: it fetches your page as a non-rendering crawler, renders the same URL in a real browser, compares the two, and grades the difference. To check whether ChatGPT can read your website, paste the URL and run a scan. It is free, needs no signup, and takes under a minute.
What is Lantad AI?
Lantad AI is Lantad: one product, the AI visibility checker at lantad.co. It measures what ChatGPT, Claude, Gemini and other AI assistants can read on your site by fetching your page the way their crawlers do, comparing that against a real browser render, and grading the gap with ranked fixes. A scan of any single page is free and needs no account.
What is generative engine optimization (GEO), and is AEO the same thing?
Generative engine optimization, or GEO, is the work of being readable and therefore citable by AI answer engines rather than only by search engines. Answer engine optimization, or AEO, is the same job under a different acronym, and the two are used interchangeably across the industry. Both overlap with SEO without being the same work: Google renders JavaScript and most AI crawlers do not, so a page can rank well and still hand a crawler nothing. GEO and AEO both start with whether the content is present at all, which is what an AI visibility checker measures.
What is Prose Parity?
Prose Parity is the share of your rendered, human-visible content that ships in the initial HTML an AI crawler reads. One hundred percent means a crawler sees everything your visitors see. A low score means content is arriving via JavaScript that most AI crawlers never run.
Which AI crawlers and assistants does Lantad check?
Lantad checks 15 named AI crawlers across OpenAI, Anthropic, Perplexity, Google, Apple, Meta, ByteDance, Common Crawl and Amazon. For each one, the report shows the robots.txt rule that applies. The 12 crawlers that publish a request user agent are also probed with that user agent, and the report shows how your server answered. The remaining 3 are robots-only tokens with no user agent to test, so the report reads their rule and says plainly that no probe was possible rather than implying one that never happened.
What is AI crawler detection, and how does it work?
AI crawler detection is finding out which AI crawlers can reach your pages. It has two halves and both are needed, because they disagree. The first reads your robots.txt and resolves the rule that applies to each named crawler token. The second sends a live request as that crawler and records what your server did, which catches a CDN rule or bot-management setting blocking a crawler your robots.txt allows. Lantad checks 15 tokens in robots.txt and probes every one of the 12 that publish a user agent of their own.
Why can AI crawlers not read my site?
AI crawlers like GPTBot, ClaudeBot, and PerplexityBot do not execute JavaScript. If your content is rendered client-side, they receive an empty shell even though the page looks fine in a browser and in Google. The usual causes are a client component that fetches its own content, a single-page-app shell with no server rendering, a storefront app injected after load, and tabs or accordions that mount their panel text on click.
Is optimising for AI the same as Google SEO?
No. Google renders JavaScript; most AI crawlers do not. A site can rank well on Google and still be invisible to ChatGPT or Perplexity. Classic SEO tools check the layer Google sees. Lantad checks the layer underneath it: how much of your content reaches crawlers that do not run JavaScript, whether your server lets them in, and whether your structure and structured data survive in the raw HTML.
Does Lantad spoof crawler identities?
The baseline fetch always identifies itself honestly as LantadBot. Per-bot probes send each bot's published user agent string once to detect blocking, with the results caveated, because servers that verify crawler IPs behave differently for the real crawler. Site owners can opt out, and the crawler conduct page documents exactly what LantadBot requests and how often.
Is the scan free, and how long does it take?
Yes. Every scan is free, needs no email and no signup, and there is no gated report to wait for. Most scans finish in under a minute. You watch the progress live and get a shareable report link at the end that anyone can open without an account.