Talk to Franz

Workshop · 1 to 2 days

Technical SEO Workshop.

Underwater, a developer hands a wooden crate with a glass side to a swimming courier; inside, a chair, a lamp, books and a picture stand fully assembled, a green tag hangs from the rope handle, and a sealed flat-pack box with instructions lies unused on the sand
AI-generated image (OpenAI, via Codex) for Full Stack Optimization

Google sees less of your site than your team thinks. In one or two days your developers learn how Google crawls, renders and indexes, on your own pages, and why each ticket matters.

For

  • Frontend, backend and fullstack developers
  • Technical product managers
  • IT and QA

Is this you?

  • Google does not index what you publish
  • JavaScript frameworks Google cannot render
  • Many URLs without page-type logic
  • SEO tickets developers do not understand

Developers decide what Google sees

Google does not see your site the way your team does. It fetches the HTML, renders it later, and indexes what it got. Google: “Google processes JavaScript web apps in three main phases” - crawling, rendering, indexing. (Google Search Central)

It works best when the first HTML is complete: “Crawling a URL and parsing the HTML response works well for classical websites or server-side rendered pages where the HTML in the HTTP response contains all content.” (source) Every one of those decisions sits in code: a redirect, a status code, a file name, a sitemap generator.

A ticket that says “fix the canonicals” does not tell a developer what to build or when to stop. A ticket with the reason, the test and a number does. In 2017 I wrote down the aim of my open developer days: make external SEO consultants, me included, unnecessary.

My goal for the day: every developer in the room can answer the next SEO question alone.

What developers get wrong in 2026

These are the failures I found while preparing my 2026 workshops, in the teams’ own Search Console data and on their live pages. For the audit side, see Technical SEO Consulting.

What goes wrongWhat I measuredWhat I specify
The right decision, the wrong status codeAbout 2k retired URLs redirected to the home page; about 5k clicks in the 90 days before410, or a one-hop 301 to a real successor
A sitemap nobody testedNone of 60 sampled sitemap URLs answered 200Only URLs that answer 200, checked on every release
A canonical read as a commandAbout 40 percent of clicks on variants the team thought Google ignoredAsk Google’s URL Inspection before any i18n change
Files of the last release are goneNone of about 90 old files still served by the origin after more than a weekHashed names, old files online for 16 months
HTML weight nobody budgetsAbout 470 KB to 1 MB per page type, against my 150 to 250 KBThe budget, a lean page, a check per release
One page, two URLsMost of about 100 ranking URLs hit a redirect to a different form of the same URLFix the source that emits the wrong form
A migration to fix a reportAlmost half of Google clicks would have gone through redirectsLabel the group in reporting, not in the URL

The right decision, the wrong status code

One team had decided that a whole group of pages was dead weight. Then it redirected about 2k of them to the home page.

  • Those URLs had earned about 5k clicks and over 1M impressions in the 90 days before.
  • A permanent redirect to the home page is the pattern Google warns about.
  • Nobody had told the team. A sample showed it.

Google: “Don’t redirect many old URLs to one irrelevant single URL destination, such as the home page of the new site.” (Google Search Central, How to move a site)

What I specify:

  • 410 for pages that will not return. No redirect in front of an error.
  • A one-hop 301 only where a real successor page exists.
  • A decision is done when the status code says so.

A sitemap nobody tested

A sitemap file listed more than a thousand URLs. I fetched 60, evenly spaced through the file.

  • None answered 200.
  • Most answered 404.
  • The rest redirected to a page that exists.

Google: “A sitemap tells search engines which pages and files you think are important in your site” (source). This one told Google the opposite.

What I specify: the sitemap lists only URLs that answer 200, are indexable and canonical to themselves. A check fetches a sample of every sitemap file on every release.

A canonical read as a command

A team feared that its translated variants would be ignored. It wanted to restructure the URLs.

I asked Google’s URL Inspection instead. Google had indexed the variants as their own URLs and ignored the declared canonical.

  • About 40 percent of the site’s Google clicks, tens of thousands in 90 days, landed on those variants.
  • Any change to the scheme is a URL migration of tens of thousands of ranking URLs.

Google calls a canonical a signal: “A strong signal that the specified URL should become canonical.” (source) A signal can lose.

What I specify: check the Google-selected canonical in URL Inspection before any i18n or canonical change. Plan a change as a migration, with a rollback.

Files of the last release are gone

I took a page capture from one release and requested every JavaScript and CSS file it referenced, more than a week later.

  • About 90 files. The origin answered 404 for all of them.
  • Minutes later a few answered 200. Those were leftover copies in one CDN edge cache, different per location and per minute.

Google may render a page with files it cached earlier: “WRS may ignore caching headers. This may lead WRS to use outdated JavaScript or CSS resources.” (source) Today’s render was fine. The risk is the window after each release.

What I specify:

  • A content hash in every file name.
  • Old files stay online for 16 months. That is my rule, not a Google number.
  • A check on every release: the files the previous HTML references still answer 200.

HTML weight nobody budgets

Every page type I measured shipped about 470 KB to 1 MB of unzipped HTML. The same inline stylesheet, about 200 KB, sat on every one. The page data embedded in the HTML alone was 220 to 650 KB. Visible text: under 1 KB to 8 KB per page.

I rebuilt one page type as plain HTML:

  • From about 650 KB to under 20 KB, with the same content.
  • Mobile PageSpeed score, 10 runs each: median from about 70 to nearly 100.

Google says to build for crawlers and users: “Keep in mind that server-side or pre-rendering is still a great idea because it makes your website faster for users and crawlers, and not all bots can run JavaScript.” (source)

What I specify: 150 KB of unzipped HTML per page, 250 KB at most. Everything that is not structural markup and content is ballast. A build fails when a page type goes over.

One page, two URLs

In a sample of about 100 URLs that rank, most answered with a redirect, and every one I could classify only normalised the URL. The three pages with the most clicks were behind that redirect.

The sitemap listed the right form. Something else still emitted the wrong one: old links, or a link generator.

What I specify: one resource, one URL, one form. The redirect is the mitigation. The fix is to find what emits the wrong form and stop it.

A migration to fix a report

A team asked: our list pages share no URL namespace, which makes analysis hard. Should we migrate them to one?

List pages were almost half of the site’s Google clicks, about 100k of about 230k in 90 days. Even my planning assumption of a 10 percent loss meant about 10k clicks per 90 days for a reporting convenience. My budget for a large URL change is a 20 to 50 percent loss.

What I specify: no migration. Label the group in the reporting layer, a classifier that costs a day and no clicks. Change URLs only for a reason that pays for the loss, and then follow Google’s site-move steps to the letter.

How the day runs

  1. Kickoff call, two to four weeks before. Your goal, your stack, which page types matter, and who joins. You send the questions your team keeps asking, in writing.
  2. I prepare on your pages. Search Console data, your live site, and the three SEO tests on one URL per page type: PageSpeed Insights, the page with JavaScript off, and Google’s live test in Search Console. The result is a test deck per page type.
  3. The day, live on your site. I do not lecture. We go cluster by cluster, in the order your questions weigh: rendering, distribution, performance, markup, links and duplicates. Every finding is reproduced against the live URL before it becomes a ticket, because a crawler is a hypothesis generator, never a verdict.
  4. IS and SHOULD specs. For each page type, the markup as it is, and as it must be.
  5. Tickets with a reason and a number. Each ends with an acceptance criterion your team can test, for example “0 URLs outside the defined namespaces at launch, verified by a crawl of staging”. Most checks belong in the pipeline you already have, as a red build.
  6. Written answers. One per question your team asked, with the evidence behind it.

In-house it runs one or two days, in English or German, for up to 12 people. Remote it splits into two half days. A 2-hour follow-up call two to eight weeks later is an option. I also ran an open Technical SEO and GEO day in December 2025.

I start every workshop with the line “SEO is 80% bullshit and 20% valuable parts!” Then we read Google’s specifications, not SEO blogs. (What Is SEO?)

What one day cannot do

After one 2026 day my notes held about 65 tasks and processes. Fewer than 10 were committed to in the room. The rest I had instructed, and nobody had said yes.

A day gives you the questions answered and the tickets written. Owners, dates and a routine come from your side. That is what the SEO Setup Workshop is for.

Where I check myself

I build my own tools, and I test them on my own work.

  • My SEO Live Test runs about 130 rules on a page. On one 2026 home page, half of its warnings were false: a HEAD request showed no compression, and a brand check failed on a spelling variant of the brand. I disproved them by hand before they became tickets.
  • The same kind of bug hit my audit tooling from the other side. It had no Brotli decoder and reported “No title tag” on pages that have one. I fixed it.
  • The tool and its code are public. The checks on this website are in the Technical SEO Consulting page.

Questions I get before a technical day

Do our developers need SEO knowledge?

No. They need their stack and their questions. The day starts from what your pages ship.

Does Google render JavaScript?

Yes, later, in a separate step. Google: “Keep in mind that server-side or pre-rendering is still a great idea because it makes your website faster for users and crawlers, and not all bots can run JavaScript.” (source) The main content belongs in the server HTML.

Can we hide content behind a tap?

Not the content you want found. Google: lazy-loading must not rely on user actions “which is important as Google Search does not interact with your page.” (source)

Do we need an llms.txt file?

Not for Google. Google: “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search” (source). A second content inventory to keep in sync is a real cost, so my written answer to one team was: not now.

One day or two?

One day if you have a short list of questions and one stack. Two if there are several page types, several teams, or a migration planned. We decide in the kickoff call.

Which stack do you need?

Any. SEO only sees the URL and the HTML that comes out the front. The stack is your decision.

What do we leave with?

The test deck per page type, the IS and SHOULD markup specs, the performance to-dos and a written answer to every question your team asked.

Sources for the Google quotes, read on 2026-11-12: Understand the JavaScript SEO basics, How to move a site, Learn about sitemaps, How to specify a canonical URL, Fix lazy-loaded content and Optimizing your website for generative AI features on Google Search, Google Search Central. The client numbers come from my own workshop preparations and handovers in 2026.

Some of our customers

Examples, not the full list

  • dpa
  • Journi
  • wikifolio
  • IMMMO
  • Dr. Schär
  • Stiftung Warentest
  • Der Pragmaticus
  • Bianor
  • RapidPipeline
  • ASI Reisen
  • WKO
  • Lendo
  • bluemonkeys
  • Tractive
  • NETCONOMY
  • Kleine Zeitung
  • Paperpile

From real workshops

  • 2026: a full technical day for a web team: 18 written questions answered with its own Search Console data, about 65 tasks and processes written down.
  • 2026: a one-day session for a product team: almost 50 written questions, one prepared answer for each.
  • 2025: a two-day workshop with sample HTML for a list page and a detail page.
  • December 2025: an open Technical SEO and GEO day.
  • 2013 to 2017: open No B*llshit SEO for Developers days in Vienna, Graz and Linz, attendees from dozens of companies.