Every number on this page was measured

§ 01 · MEASUREMENT DETAILS · 2026-07-30

Which tools produced these values, and under what conditions?

Which rules does the method follow?

• The measured address is written on every row. Different pages produce different results; we do not claim that a single score represents the whole site.
• The published set of pages is not selected. Every indexable address in the sitemap is measured and every one is printed; the table starts with the lowest-scoring page. Publishing the good pages out of the ones you measured is cherry-picking, even when every row is individually correct.
• Lab measurement is taken on the mobile profile. The desktop score is almost always higher, which is why it is not published here.
• Values are not rounded up. The format the tool returns is preserved; a score of 97 is not presented as 100.
• A measurement that cannot be repeated is not published. If a value falls in the next round, the lower value is written and the row stays.

This is the same measurement and acceptance-criteria discipline we apply to client projects. The full working model is written out on the How We Work page.

§ 03 · CORE WEB VITALS

What are the Core Web Vitals values, in the lab and in the field?

Lab values

• Largest Contentful Paint (LCP) 1.4s Lighthouse 13.4.1 · mobile lab · 2026-07-30 verify ↗
• Cumulative Layout Shift (CLS) 0.00 Lighthouse 13.4.1 · mobile lab · 2026-07-30 verify ↗

Field (CrUX) data

What thresholds has Google published?

The thresholds are taken from Google's public documentation. For the definition of each metric and how it is fixed, see the Core Web Vitals guide; for the scope of the work itself, see Core Web Vitals and performance optimisation.

§ 04 · STRUCTURED DATA

Is the structured data on this page valid?

• Nodes in this page's JSON-LD graph 8 Published HTML · 2026-07-30
• ld+json tags per page 1 Published HTML · 2026-07-30

Which schema types does the site use?

The shared graph is identical on every page: Organization, ProfessionalService, Person (the founder), WebSite, WebPage and, everywhere except the home page, BreadcrumbList. Page-type nodes are added on top: Service and HowTo on service pages, TechArticle and DefinedTerm in the knowledge centre, ItemList and CollectionPage on hub pages, and FAQPage on every page that carries questions and answers.

{
  "@context": "https://schema.org",
  "@graph": [
    { "@type": "Organization",       "@id": "https://www.bakibilisim.com/#organization" },
    { "@type": "ProfessionalService", "@id": "https://www.bakibilisim.com/#localbusiness" },
    { "@type": "Person",             "@id": "https://www.bakibilisim.com/#erdibaki" },
    { "@type": "WebSite",            "@id": "https://www.bakibilisim.com/#website" },
    { "@type": "WebPage",            "@id": "https://www.bakibilisim.com/en/proof/#webpage" },
    { "@type": "BreadcrumbList",     "@id": "https://www.bakibilisim.com/en/proof/#breadcrumb" },
    { "@type": "HowTo",              "@id": "https://www.bakibilisim.com/en/proof/#howto" },
    { "@type": "FAQPage",            "@id": "https://www.bakibilisim.com/en/proof/#faq" }
  ]
}

All @id values are consistently https and www; nodes are linked to each other by identity rather than copied. The method is explained in the Schema.org guide and its delivery scope on the structured data service page.

§ 05 · ACCESSIBILITY

Against which criteria was accessibility checked?

Measured contrast ratios

What was checked by hand

• Every interaction can be completed with the keyboard; focus order follows DOM order and there is no keyboard trap.
• Every focusable element has a visible focus ring (:focus-visible, 2px, 2px offset).
• The first Tab press reveals the Skip to content link; the main content is marked with <main id="main">.
• Buttons, link list items and accordion headers have a touch target of at least 44×44 pixels.
• Each page has a single h1, heading levels are not skipped, and the landmarks (header, nav, main, footer) are defined.
• Icon-only buttons carry an aria-label and decorative SVGs aria-hidden; horizontally scrolling tables and code blocks are reachable by keyboard through role="region" and tabindex="0".
• With prefers-reduced-motion enabled all transitions are stopped; nothing autoplays, scrolls or counts on its own.

Automated tools cannot check every WCAG success criterion; criteria that depend on meaning, order and context can only be verified by hand. That is why we publish the automated score and the manual checklist together.

§ 06 · WEIGHT AND REQUESTS

How many kilobytes and how many requests does this page cost?

• First-load transfer weight of this page (compressed) 103.8KB Network log + build report · 2026-07-30
• File requests on first load 8 Network log · 2026-07-30
• Third-party requests on first load 0 Network log · 2026-07-30
• Total JavaScript weight 4.0KB gzip Build report · 2026-07-30

The table reports the transfer sizes Lighthouse recorded: the brotli q11 compressed body plus the HTTP response headers (CSP included, roughly 0.4 KB per response). The body-only total is 100.5 KB. Because each row is rounded from a byte measurement to one decimal, the column may not add up to the measured total exactly; the total row is the measured byte total, not the sum of the rounded rows. One caveat we owe you: this row is part of the page it measures, so writing the figure moves the result by a few dozen bytes. The font files are cached after the first visit, so subsequent page transitions download only the HTML.

§ 07 · AI CRAWLER ACCESS

Can AI crawlers reach this site?

• Explicitly allowed AI crawlers 20 robots.txt · 2026-07-30 open file ↗
• llms.txt file size 6.0KB Source file · 2026-07-30 open file ↗
• Addresses listed in sitemap.xml 72 Build output · 2026-07-30 open file ↗

Which crawlers are allowed?

GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, Claude-Web, Claude-SearchBot, anthropic-ai, PerplexityBot, Perplexity-User, Google-Extended, Google-CloudVertexBot, CCBot, Applebot-Extended, Amazonbot, Bytespider, meta-externalagent, MistralAI-User, cohere-ai, YouBot and DuckAssistBot. 7 classic search engine crawlers (Googlebot, Googlebot-Image, Bingbot, DuckDuckBot, YandexBot, Applebot, Slurp) are named and allowed as well. The general rule (User-agent: *) also grants full access; no directory is blocked.

This is a written decision rather than a default: our objective is to be cited in answer engines and generative engines. The decision belongs to the client and is documented as a comment inside the robots.txt file; for an organisation that wants access closed, the same file can be written the other way round.

Why and how this policy is set up is explained in the What is GEO and What is AEO guides.

§ 08 · COMPARISON · ANONYMISED

Where do these values sit within the sector?

Method: the same tool and the same mobile profile, home page address only, uncached first load, transfer size. The sample is 5 sites; no claim of statistical representativeness is made and it is not presented as a sector average. We do not publish names because those measurements were not taken with the other party's consent — our own measurement, by contrast, is published together with its address.

What did the previous version of this domain measure?

The WordPress version this site replaced was measured with the same Lighthouse build and the same mobile profile as the new one. Since it is our own history, nothing is anonymised in this row.

The previous version was measured on 29 July 2026 and the new version on 30 July 2026, with the same Lighthouse 13.4.1 build and the same mobile profile. The previous version was measured on the live address; the new version was measured on the production output (dist/) served under the same compression and header policy. Weight and request counts are directly comparable; because the two runs sat at different CPU benchmark indices (1314 and a median of 3924), the CPU-bound rows — TBT, TTI and the performance score — are not strictly equivalent: the direction and the order of magnitude hold, decimal equality does not. The previous version ran on WordPress and Elementor, and a robots meta tag kept it out of search engine indexes entirely.

§ 09 · MEASUREMENT HISTORY

How is the measurement history kept?

The weight column is the first-load transfer size of the measured address. When the next round is published a new row is added to this table and the existing rows are left untouched.

§ 10 · VERIFICATION

How can you verify these scores yourself?

1. 01

   Measure the Lighthouse scores with PageSpeed Insights

   Enter this site's address in the address field and read the mobile tab of the report. PageSpeed Insights runs the same Lighthouse engine on Google's servers, so the result is independent of your device and your network.

   Tool

   pagespeed.web.dev ↗ (opens in a new tab)

   Expected outcome

   Four category scores plus lab LCP and CLS values; compare them with the § 02 and § 03 rows on this page.

2. 02

   Validate the structured data

   Paste the page address into the Schema.org validator: it lists syntax and type errors. Google's Rich Results Test reports the same graph separately, for rich result eligibility. The two tools look at different things, so both need to be run.

   Tools

   validator.schema.org ↗ (opens in a new tab) · Rich Results Test ↗ (opens in a new tab)

   Expected outcome

   Zero errors; a single ld+json tag with nodes linked to each other by @id.

3. 03

   Scan for accessibility and navigate by keyboard

   Scan the page with WAVE or axe DevTools. Then let go of the mouse and move through it with the Tab key only: the first press should reveal the Skip to content link, every focused element should show a visible ring, and at no point should you fall into a keyboard trap.

   Tool

   wave.webaim.org ↗ (opens in a new tab) or axe DevTools as a browser extension

   Expected outcome

   No contrast errors; a single h1; landmarks defined; every interaction reachable by keyboard.

4. 04

   Count weight and third-party requests in the network log

   Open the Network tab in developer tools, disable the cache and reload the page. Look at the domains in the request list: you should see no domain other than bakibilisim.com. The total transfer size at the bottom can be compared with the § 06 table.

   Tool

   Browser developer tools — Network tab

   Expected outcome

   Zero third-party domains; a request list containing the HTML, one CSS file, one JS file, two fonts and the small files the browser asks for (site.webmanifest, favicon.svg) — 8 requests in total. favicon.svg is requested twice; the second one is served from cache and transfers 0 bytes but still counts as a request.

Your result may differ from what we publish: network, device, browser extensions and the moment of measurement all change the outcome. If you find a significant difference, write to bilgi@bakibilisim.com with a screenshot; we will look into it and update this page.

The free digital asset audit covers part of these steps on your own site: repeated Lighthouse measurements on a mobile profile, a scripted structured data rule check, an automated axe-core accessibility scan and the transfer size of the home page, with the findings reported item by item. It does not include the manual keyboard checks.

§ 11 · QUESTIONS

Frequently asked questions about the measurements

The full proof architecture — measurement plus an open methodology, a transparent contract and an agency selection checklist — is on the Why Baki Bilisim page. Every other question is answered on the FAQ page.