A real audit · nothing removed

We ran the audit on ourselves and published what it found.

Every number below was measured by the same twenty-seven agents a paying client gets. We picked our own site because we cannot publish a client’s audit without their permission, and we would rather show you real output than a mock-up.

Site audited: quintarthai.com Run: 29 August 2026 Pages read: 25 Agents run: 23

What the instruments read

Raw measurements, before anyone interprets them. Nothing here is an estimate.

LCP
1,836 ms
real Chromium, cold load
Server response
523 ms
home page, single request
HTML weight
393 KB
402,588 bytes, gzip on
Pages crawled
25
of 404 in the sitemap

Findings

Ordered by what costs you most, not by which agent found them. Each one carries the raw observation it came from, so you can check it yourself before you act on it.

Critical · email deliverability

The domain publishes no MX record. It cannot receive email.

domain quintarthai.com MX not_found SPF found (v=spf1, 1 DNS lookup, within the RFC 7208 limit of 10) DMARC found DKIM no record at any of 9 selectors tried default, google, k1, s1, s2, selector1, selector2, mail, dkim

SPF and DMARC are both published and correct, so outbound mail is authenticated. But with no MX record, anything sent to this domain has nowhere to land — including reply-to addresses, password resets and anything a prospect types into a contact form.

Fix: publish an MX record pointing at your mail host. If the domain is deliberately send-only, publish MX 0 . (the null MX, RFC 7505) so senders fail fast and immediately instead of silently queuing for days.

On DKIM: a selector cannot be enumerated from DNS. “No record at the nine selectors we tried” is not the same as “DKIM is absent”, and we will not write it up as though it were.

High · conversion

Three of six pages read offer the visitor nothing to do.

pages fetched 6 pages failed 0 pages with zero CTAs 3 https://quintarthai.com/engines.html https://quintarthai.com/engines https://quintarthai.com/methodology/composite-score.html forms over field cap 0

These are explainer pages that do the persuading and then stop. A reader convinced by the methodology page has no next step from it.

Fix: one clear action at the foot of each. Not a nav link — a specific next step that matches what the page just argued.

What we will not tell you: that this will lift conversion by some percentage. We do not run your A/B tests, so we do not have that number and will not invent one.

Medium · page speed

393 KB of HTML on the home page, served with no-cache.

LCP 1836.0 ms (headless chromium, one cold load, desktop viewport, unthrottled network) DOMContentLoaded 1163.8 ms CLS no-layout-shift-entries-observed response 522.8 ms html_bytes 402,588 content-encoding gzip cache-control no-cache, must-revalidate

LCP at 1.84s is inside Google’s 2.5s “good” threshold, so this is a headroom problem, not a failure. But 393 KB of HTML is roughly ten times a typical marketing page, and no-cache means every repeat visitor re-downloads all of it.

Fix: serve static HTML with a real max-age and revalidate with an ETag. Move whatever is inlined into a cacheable file.

Scope, stated: this is a server-side measurement from one machine on an unthrottled connection. It is not field data from your actual visitors, and CLS reporting “no entries observed” means exactly that — not that CLS is zero.

Medium · accessibility

Four of six pages have no skip link, and one skips a heading level.

pages audited 6 skip_link_missing 4 headings_skipped_levels 1 images_missing_alt 0 html_lang_missing 0 form_inputs_unlabeled 0 links_generic_text 0 buttons_no_accessible_name 0 tables_without_th 0 positive_tabindex 0 iframes_without_title 0

Most of this page set is clean. A keyboard user currently tabs through the whole navigation on every page before reaching content, and one page jumps a heading level, which breaks the document outline a screen reader announces.

Fix: add a visually-hidden skip link as the first focusable element, and correct the one heading jump.

Scope, stated: these are markup checks. Colour contrast, keyboard operation and screen-reader behaviour were not tested here, and this is not a complete WCAG audit. In Ontario, accessibility is a statutory obligation under the AODA, not an optimisation — this is general information, not legal advice.

Clean · structured data

Schema is present, parses, and claims nothing false.

pages analyzed 9 types present FAQPage, Organization, SoftwareApplication, WebSite parse errors 0 required-property gaps 2 rating markup observed false

Nine pages carry valid JSON-LD with no parse errors. Two required properties are missing on otherwise-correct blocks. Notably there is no rating or review markup — which is correct, because there are no reviews to mark up. Fabricated AggregateRating is a penalty and a lie about a real business, and our generator strips it rather than emit it.

Fix: fill the two missing properties. Nothing else here needs touching.

The measurement we refused to report

This is the part most audits leave out, and it is the reason to trust the rest.

Orphan pages: not reported.

pages_crawled 25 page_cap 25 cap_hit true orphan_count { "available": false, "reason": "crawl stopped at its 25-page cap after 25 page(s) while the sitemap lists 404; an orphan cannot be told apart from a page the crawl never reached" }

An orphan is a page nothing links to. Our crawl stopped at 25 pages; the sitemap lists 404. Every one of the 353 pages we did not open looks unlinked from where we stopped.

An earlier version of this agent reported that number as “353 orphan pages”. It was wrong, and it was wrong in the direction that manufactures alarming findings — the kind a client pays to fix and then discovers were never broken. We found it, fixed it, and the engine now refuses the number instead of guessing it.

You will see refusals like this in your report. Where a measurement is not available, we say so and say why. A number we cannot stand behind is worth less than an honest blank.

What this audit did not cover

Stated up front, because a limit you find out about later reads as a lie.

25 of 404 pages. The link-graph crawl hit its cap. A larger site gets a stated cap and the arithmetic that goes with it, never an implied “we read everything”.

No field performance data. Lab measurement from one machine, not your real visitors on real devices.

No colour-contrast or screen-reader testing. Markup checks only.

No ranking forecast. We do not predict traffic and do not promise positions. Nobody controls a search index.

Your site, the same way, for $149.

One flat fee. No retainer, no call required. If you read the report and find nothing you can act on, tell us within 7 days and we refund all of it.

Order the audit — $149