Katarzyna

Katarzyna

@katrzyna.spinalcms.com

Doing marketing (and sales-related tasks) at Spinal CMS.

17 followers120 following89 posts

Search measurement used to be one line: rank, click, convert. Yay! AI search broke it. Boo! The answer now arrives before the click (and often without one). Your analytics never see that step.

View full thread

Keep reports true. Split them by engine. Read patterns across many prompts, not one. Separate AI-referred from AI-influenced outcomes. Treat any vendor score as a sample, not a fact.

Search measurement used to be one line: rank, click, convert. Yay! AI search broke it. Boo! The answer now arrives before the click (and often without one). Your analytics never see that step.

The old chain was already weak: privacy changes limited the data access. AI made it worse. Last-touch attribution reads the return visit as direct or branded traffic. Then it tells you: “AI did nothing”.

Do not chase one number, but measure three layers: discovery, recognition and outcome. Can bots reach you? Do you appear in answers? Do people buy? Fix discovery first.

The features coming to Spinal are really exciting for (small) marketing teams. Things are going to get so much easier to get done faster and better. That is all I can say, right now! 😅

Coming to a spinalcms.com team near you soon: invite a hosted or BYOK #LLM / #ai agent member. We are testing this internally and it is great way to leave the tedious and repetitious task to do it for me.

Katarzyna reposted

The pitch for single page apps was always “it’s faster after the first load”. Google’s Core Web Vitals, used in search ranking, now measure soft navigations. It’s true: LCP drops by about a third. But, I don't think SPAs are ever really worth all the compromises and extra code.

One the coolest (new) features from spinalcms.com is templates ✨ It give you a ready-made starting point for new content. Each collection has a default template that pre-fills body (and frontmatter) of every new resource. Set a default for every new resource or select one by command/right click.

View full thread

Git-based route: connect your repo. It reads the docs folder, infers the fields and commits your edits to Git. Astro builds the same files as before. Three steps, no upkeep.

API route: build a content model, import the existing pages, manage API tokens, remap Starlight's frontmatter fields, move images to a CDN, add rebuild webhooks. Ten steps, then upkeep.

Katarzyna reposted

Zero-click does not mean zero-website. Impressions climb while visits fall. Your site now works before anyone arrives. It feeds AI Overviews, ChatGPT answers, and Perplexity citations. I do not make the rules. 🤷 For your static site, that is not a threat. It is your existing workflow.

Zero-click does not mean zero-website. Impressions climb while visits fall. Your site now works before anyone arrives. It feeds AI Overviews, ChatGPT answers, and Perplexity citations. I do not make the rules. 🤷 For your static site, that is not a threat. It is your existing workflow.

View full thread

AI answers are bland by nature, so your weird and human voice wins. That voice lives on the site you fully own: the fonts, the words, the strange phrase that says “not a bot”.

Zero-click does not mean zero-website. Impressions climb while visits fall. Your site now works before anyone arrives. It feeds AI Overviews, ChatGPT answers, and Perplexity citations. I do not make the rules. 🤷 For your static site, that is not a threat. It is your existing workflow.

View full thread

llms.txt is one Markdown file at your root. It tells AI tools what you are, in their terms. llms-full.txt inlines every page, so an AI tool reads your whole site in one fetch. On a static site generator, you build it from frontmatter you already write. About twenty lines. Minutes of work!

The risk: an AI tool quotes an outdated version of you. On a static site, exactly one version exists. You fix the file, then rebuild. That build is what gets crawled tomorrow. No cache to expire. No draft that forgot to publish.

Zero-click does not mean zero-website. Impressions climb while visits fall. Your site now works before anyone arrives. It feeds AI Overviews, ChatGPT answers, and Perplexity citations. I do not make the rules. 🤷 For your static site, that is not a threat. It is your existing workflow.

You already have a home base. Your content is a folder of Markdown in Git. That is one true source you own. Not a database row behind a login, nor a post on some platform. “Write once, build everywhere” is what zero-click now demands of everyone else.

Katarzyna reposted

SEO on a static SaaS site still works. But AI search changed where buyers start. 51% of B2B buyers now begin research in a chatbot more often than in Google. Ranking first no longer decides it. Getting cited inside the AI answer does.

SEO on a static SaaS site still works. But AI search changed where buyers start. 51% of B2B buyers now begin research in a chatbot more often than in Google. Ranking first no longer decides it. Getting cited inside the AI answer does.

View full thread

SEO on a static SaaS site still works. But AI search changed where buyers start. 51% of B2B buyers now begin research in a chatbot more often than in Google. Ranking first no longer decides it. Getting cited inside the AI answer does.

View full thread

A static stack gives you speed, control, and build-time templates for free. Core Web Vitals pass by default. What you hand-roll instead: JSON-LD, headings, robots.txt.

Three cheap moves: ship JSON-LD from one partial, write headings as questions, and allow AI search bots in robots.txt while blocking training crawlers.

SEO on a static SaaS site still works. But AI search changed where buyers start. 51% of B2B buyers now begin research in a chatbot more often than in Google. Ranking first no longer decides it. Getting cited inside the AI answer does.

Google's number one result fell from 28% to 19% click-through rate after AI Overviews shipped. Two goals now: rank in the classic results, and get quoted by the chatbot.

Katarzyna reposted

Google now lets readers pick your site as a "preferred source." Those readers then see your content flagged with a badge in Top Stories, AI Mode and AI Overviews. It is optional, and it asks something of you first. My thoughts.

Google now lets readers pick your site as a "preferred source." Those readers then see your content flagged with a badge in Top Stories, AI Mode and AI Overviews. It is optional, and it asks something of you first. My thoughts.

View full thread

My take? If you want the badge, start with the plain deeplink. It costs nothing and needs no script. Skip the JavaScript. The readers worth keeping come to your domain with or without a badge.

Google now lets readers pick your site as a "preferred source." Those readers then see your content flagged with a badge in Top Stories, AI Mode and AI Overviews. It is optional, and it asks something of you first. My thoughts.

View full thread

The reward? A “preferred“ badge inside AI surfaces that answer without a click. You would be promoting a Google feature to stand out inside the same system that reduced your traffic. 🤡

Also worth remembering: Google's pattern with publishers. - AMP asked for a parallel site format - Authorship asked for markup - Web Stories asked for a new page type. All three took real effort. All three were later retired. Just a pattern…

Google now lets readers pick your site as a "preferred source." Those readers then see your content flagged with a badge in Top Stories, AI Mode and AI Overviews. It is optional, and it asks something of you first. My thoughts.

Google “recommends” loading its JS on every page of your site to render the button. But the preference only matters inside Google. The data flows to them. The work flows to you. 🫠

Katarzyna reposted

AI tools cite other websites when they describe your company. A study of 1,851 sources found brand-owned pages made up 2.8% of citations. You can fix that.