Send us your site. We find the structured data and alt text that machines can’t read, and fix a slice free.

We audit up to 100 of your public pages for two defects, and fix them as drop-in files. Nothing to install, no account, no access to your site. The sample is free. If you want the rest done, you get a fixed price, quoted after the sample — never before.

Request the free audit →

Opens your own mail client with a short template. No mail client? Write to lilu@ability.ai with the subject “Free audit request”.

The two defects we fix

  1. Missing or broken structured data (ld+json) on listing or product pages. You get a JSON-LD block per page, ready to drop in.
  2. Missing image alt text on a product catalogue. You get an import-ready CSV.

Those two only. They are the two we have actually performed and verified, so they are the two we offer.

What it looked like on real sites

Measured 2026-10-03. Neither site is named: neither agreed to be a case study.

On a live property-listing site, 100 pages sampled: 90 of 100 received a schema-valid JSON-LD block. Every field the page did not state was left null rather than invented.

On a live fashion retailer’s catalogue, 80 product rows: 80 distinct alt-text strings, none empty, longest 106 characters, delivered as a CSV that imports through the store’s own product importer with nothing installed.

We ran it on our own site first

Fetched 2026-10-04, 17:02:48–17:05:00 UTC. 55 HTTP requests, no login, one request at a time, at least two seconds apart. Every number in this section was derived from those saved bytes in that run, and nothing here is carried over from an earlier one.

We are asking you to let us read your pages, so here is the same check run against every page the lab itself publishes — six hosts, 23 pages, defects included and not a flattering subset. The 23 are every URL the hosts’ own sitemaps and front pages declared; three of the six hosts are one page each, and two of those are closed-project notices, so the pages are not spread evenly.

Our own six hosts, audited 2026-10-04.
HostPagesNo JSON-LDAlt-text defects
lilulab.ai110
ninemin.lilulab.ai18180
willcall.lilulab.ai110
ghmirror.lilulab.ai110
audit.lilulab.ai110
pinpoint.lilulab.ai110
All six23230

Defect 1 — found on 23 of 23 pages. Not one page the lab publishes carried an ld+json block at all. That is the first and bluntest of the two things we look for: not markup that is subtly wrong, but markup that is absent. Every host failed it, including this one.

Defect 2 — 0 defects, and the dull reason matters. Across all 23 pages there was exactly one <img> element in total, and it already had a non-empty, non-placeholder alt. So this class came back clean because there was almost nothing on our own site to check — not because we are good at it. A text-only site is the easy case, and we will not count it as evidence we can fix yours. The alt-text work we have actually performed was on a product catalogue, not here.

Found 23, fixed 21. This page was the first one we fixed, below; 20 more were fixed the same way, from each page’s own bytes, and every field a page did not state was left out rather than invented. The two we did not fix are shut projects carrying noindex, where markup is inert — we skipped them rather than padding the count. You can check all of that too.

The fix, on this page, which you can check yourself

This is the whole of what you receive: the broken region, and a block to drop in. Nothing to install, no build step, no access to your site.

Before — https://audit.lilulab.ai/ as served at 17:04:57 UTC, 6957 bytes. The head goes straight from the canonical link to the stylesheet; there is no ld+json anywhere in the document.

<link rel="canonical" href="https://audit.lilulab.ai/">
<!-- No script, no form, no cookie, no external request. The only intake is
     a mailto: link that opens the visitor's own mail client. -->

After — the same region with the drop-in block in place. This is live: open this page’s source and you will find exactly these bytes.

<link rel="canonical" href="https://audit.lilulab.ai/">
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Service",
  "name": "Free two-defect page audit",
  "serviceType": "Structured-data and image alt-text audit",
  "url": "https://audit.lilulab.ai/",
  "description": "We audit up to 100 of your public pages for missing or broken JSON-LD and missing image alt text.",
  "provider": {
    "@type": "Organization",
    "name": "Lilu Lab",
    "url": "https://lilulab.ai/"
  }
}
</script>
<!-- No executable script, no form, no cookie, no external request. The only intake is
     a mailto: link that opens the visitor's own mail client. -->

The same fix written out end to end, on a page where both halves can be checked — this lab’s own front page, measured with no markup at 18:06 UTC and with it at 18:58 UTC the same day, and the two pages still carrying the defect at 19:00 UTC — is at no structured data on the page.

Note what the block does not say. There is no price in it, because the price is not on this page either — it is quoted after the sample. There is no areaServed, no rating and no review count, because this page does not state any. A field the page does not support is left out rather than filled in with something plausible. That rule is the reason the files are safe to drop in unread, and it is the rule we apply to your pages too.

How to re-run this and catch us out

If you cannot reproduce a result, you are reading marketing. So:

  1. Fetch the page yourself — curl -s https://audit.lilulab.ai/ — or just use your browser’s view-source. No key, no account; these are public pages.
  2. Search the bytes for application/ld+json. Zero hits is defect 1, case one. That is the entire test for the case we found on all 23 of our pages.
  3. If there is a hit, paste what is between the tags into any JSON parser. If it does not parse, that is case two.
  4. For alt text: find every <img and read its alt. Missing attribute, empty string, or a filler word like “image” or “photo” is a defect; anything else is not.
  5. Our hosts and the fetch times are in the table above. A page may have changed since — ours certainly did, because we fixed this one.

What this check does not cover. Being clear about the edges is the difference between an audit and a claim:

How it works

  1. You email your site URL using the button above. That is the whole intake.
  2. We read up to 100 of your public pages — the same pages any visitor can see.
  3. You receive the defects we found, and a slice of them fixed as drop-in files.
  4. If you want the rest fixed, we quote one fixed price. You can stop at the free sample.

Who does the work

The audit and the fixes are produced by an automated system. No result will be sent to you without a person reading it first.

Request the free audit →

Or write to lilu@ability.ai.