TL;DR
- The hype surrounding AI suggests full automation, but the reality is quite different.
- We’ve been using AI tools daily in our actual work for over two years.
- Claude helps with drafting and structuring text, but it doesn’t replace your own ideas.
- Benefit: an honest assessment of where AI saves time and where it doesn’t.
In a nutshell:
- On TikTok and LinkedIn, it currently looks as if everyone is switching their companies over to AI.
- Claude instead of employees.
- Agents instead of processes. 400 pieces of content per week, automatically. We’ve been using AI tools daily in our work for over two years. Here’s what we’ve actually experienced. Why we’re writing this The AI hype is loud.
On TikTok and LinkedIn, it currently looks as though everyone is switching their business to AI. Claude instead of employees. Agents instead of processes. 400 pieces of content per week, automatically. We’ve been using AI tools daily in our work for over two years. Here’s what we’ve actually experienced.
Why we’re writing this
The AI hype is loud. That makes it hard to distinguish real insights from sales pitches.
We are not AI researchers. We are web developers and SEO consultants working with a dedicated team at Waterproof Web Wizard GmbH, where we maintain websites, plan content, and solve technical problems every day. We use AI tools where they save us real time. And we leave them out where they create more work than they save.
This is not a product comparison or a ranking. It’s our day-to-day work.
What we use every day
Claude for text drafts and structuring
Claude (by Anthropic) has been our primary AI tool for writing since early 2025. We use it for:
First drafts of blog posts. We enter the context, key information, and our own assessment. The draft comes back. We revise everything. What remains sounds like us. What doesn’t fit gets cut. The time saved comes from overcoming writer’s block.
Structuring content. We describe a topic and ask Claude to provide us with three different outline suggestions. Then we choose what fits our argument. That takes two minutes. Manually, it takes 20 minutes.
Technical documentation. When we explain a TYPO3 configuration or a WordPress filter, we write the code and ask Claude to formulate the explanatory text for it. Then we proofread it. This works well because we can evaluate the result from a technical standpoint.
What Claude can’t do: have its own experiences. Anything based on our projects, our clients, or our own observations, we have to write ourselves. Claude doesn’t know our projects. It synthesizes what’s on the web. If we publish that without adding our own substance, it’s a summary of summaries. Google now finds that pretty uninteresting.
Search Console and DataForSEO for Data Analysis
We check weekly which pages of our client projects are losing or gaining visibility. These are figures from Google Search Console and DataForSEO. AI helps us interpret this data more quickly.
Specifically: We export the keyword data, upload it to Claude, and ask for patterns. Which pages are losing ground in informational search queries? Which keywords are searched more frequently but get fewer clicks? This gives us an initial assessment in just a few minutes, which we then verify manually.
This doesn’t replace expertise. It speeds up the process of diving into a large volume of data.
AI for Code Research
We’ve been writing code for TYPO3 and WordPress for 18 years. Still, when we need a TypoScript configuration block or a WordPress filter hook example, we ask Claude first. Not because we don’t know it, but because the exact syntax comes faster from an AI response than from the documentation.
The key point here: We understand what the code does. We don’t copy code we can’t read. That’s not an AI tip. It’s a basic rule of software development.
What we’ve tried but aren’t using anymore
Automatic content generation for client websites
We tested it: placing AI-generated text directly on client websites. The result was technically correct and linguistically fluent. But it lacked what matters on B2B websites: original substance, concrete service descriptions, specific projects.
Furthermore: Google is getting increasingly better at detecting machine-generated content with no added value. In early 2026, we analyzed several websites that had published massive amounts of AI-generated content. The visibility trends were sobering.
For informational blog posts with a unique perspective, AI is a useful tool. For mass-produced content on service pages, it is a risky investment.
AI agents for automated tasks
We tested various agent setups. The idea: The agent performs several steps without us having to manually trigger each one.
The reality: Agents make mistakes. Not because they’re bad, but because they take instructions literally and don’t consider context. A single wrong step in an automated workflow can produce results we don’t notice until hours later.
A colleague from the Ads department told us that automated Meta Ads agents blocked entire ad accounts without clear rules. That’s an extreme example. But it illustrates the fundamental problem: Anyone who unleashes an agent on a system they don’t fully understand pays the price for the mistakes.
We use AI support for steps where we can immediately verify the result. No autonomous actions on systems that involve customer data.
What this means for SMEs
We regularly speak with business leaders in manufacturing, medical technology, and services in the Lake Constance region. The topic of AI comes up more and more often.
Our honest assessment:
AI is no substitute for expertise. If you don’t know your subject matter, you can’t evaluate the results of an AI. Errors aren’t immediately apparent. They become apparent when a customer reports them or Google penalizes the site.
The benefit lies in routine tasks: draft texts, summaries, initial research, code snippets. Here, AI saves time. When it comes to strategic decisions, client consulting, and interpreting data from your own operations, it cannot replace experience.
Switching tools takes time. It took us several months to set up our AI workflow so that it truly helps us. Anyone who thinks they can buy a tool and be immediately more productive is underestimating the learning curve.
Claude vs. ChatGPT: We mainly use Claude because we prefer its writing style and find the long context windows useful for document analysis. ChatGPT remains strong for coding tasks. For most SMEs, the difference doesn’t matter in day-to-day operations. What’s more important is to just get started and define a specific use case.
How we use AI tools at Waterproof Web Wizard
We use Claude every day. We use it for drafting text, structuring content, explaining code, and interpreting large amounts of data.
What we don’t do: deploy AI on client systems without vetting it first. Every blog post that appears on a client’s website is read and revised by us before it goes live. Every line of code that lands on a production system is understood and tested by us.
This isn’t a lack of trust in AI. It’s a craftsman’s mindset. If you want to deliver good work, you test your tools.
When clients ask us if they should use AI tools for their website, we ask back: For what exactly? The answer to this question determines whether AI helps or creates work.
Frequently Asked Questions
Is Claude better than ChatGPT for corporate websites?
That depends on the use case. In our experience, Claude has a slight edge in terms of flow and structure for longer documents and text drafts. ChatGPT remains strong for coding tasks and technical explanations. For most SME applications, the difference is negligible. It’s more important to define a clear use case and test the tool in that context.
Will our website lose rankings if we publish AI-generated text?
AI-generated text isn’t a ranking issue as long as it has its own substance and has been editorially revised. Google evaluates content based on depth, relevance, and trustworthiness, not on the method of production. It becomes problematic when text is published en masse without adding value. Those who use AI as a tool rather than a substitute for their own expertise will have no problem.
How can we, as an SME, get started with AI tools without a lot of effort?
A specific use case is better than a broad introduction. Our suggestion: First, write the text for your next customer email yourself in bullet points, enter these bullet points into Claude or ChatGPT, read the result, and revise it. This takes 20 minutes and shows you where AI helps and where it still requires work. Then you can decide how to proceed.
How much does Claude cost for businesses?
Anthropic offers Claude at various tiers. The Pro plan currently costs around $20 per month and provides access to the most powerful model (Opus). Package prices are available for teams. For most SME use cases, the Pro plan is sufficient.
Want to know if and how AI tools fit into your web project? We’ll take a look at it together with you. No hype, no sales pitch. Request an initial consultation.
Dennis Hüttner Waterproof Web Wizard GmbH
Sources
- Anthropic: Claude for Work (Enterprise Plug-ins and Cowork Integrations, February 2026) anthropic.com
- CIO: Anthropic targets core business systems with new Claude plug-ins (February 25, 2026) cio.com
- The AI Corner: Everything Claude Has Shipped in 2026 (March 28, 2026) the-ai-corner.com
- SISTRIX: AI Overviews in Germany: How Much Click-Through Rates Are Really Dropping (February 2026) sistrix.de