👋 Welcome back to AI for SME Success, your weekly dose of practical AI insights that matter to small businesses.
Today we cover:
- How to take advantage of Google’s “Preferred” label in AI Mode
- What to watch for when using AI for coding and customer support
- The best AI tools and workflows for spreadsheet automation
Let’s dive in! 👇
⭐ How to Claim “Preferred” Label in Google’s AI Mode
Google is cracking down on AI search spam and rewarding trusted websites.
In March we discussed why AI search quality lags so far behind Google search. We shared an experiment by a technology reporter who tricked AI tools by publishing a spammy, well-optimised web posting. “These AI tricks are so basic they’re reminiscent of the early 2000s, before Google had even introduced a web spam team,” said the reporter. Ahrefs’ study of 26,283 source URLs reiterated the finding: one third of ChatGPT’s favourite sources are spammy enough to be invisible on Google.
- First, Google opened the Preferred Source label, previously limited to news outlets, to any website publishing fresh content.
- Second, Google is rolling out a Highly Cited badge for articles that other publications referenced frequently.
What this means for you:
- Website owners can now designate their sites as a preferred source, pushing fresh content directly into Google’s AI Overviews and AI Mode results.
- Internet users are getting more reliable content. When you mark a website as a “Preferred Source,” Google adapts its system to prioritize that domain.
- Confirm your domain is eligible by entering your site URL directly at google.com/preferences/source. If it appears, you’re eligible.
- Add a link to your social posts or site:
- Add a Google-provided “Preferred Source” button alongside your existing social CTAs.
Once a reader clicks and adds your site, your content appears with a “Preferred” badge in their Top Stories, AI Overviews, and AI Mode.
Note. The Preferred Sources tool rollout is very recent (May 27, 2026). Not all eligible sites are populated yet. If your site is not there yet, check again in a few weeks.
🔓 Most AI-Coded Apps Have Security Gaps
Prototype with AI. Never skip the security review.
A growing body of research suggests AI-generated code often works but is not secure. A widely cited 2026 audit found that 91.5% of vibe-coded applications contained at least one AI hallucination-related security flaw. This does not mean 91.5% of code is vulnerable. It means most applications had some significant weaknesses.
The finding is supported by several independent studies:
- Veracode’s Spring 2026 analysis of 150+ AI models found only 55% of AI coding tasks produced secure code.
- Georgia Tech’s Vibe Security Radar confirmed a growing number of new AI-related vulnerabilities.
- Escape’s scan of 5,600 production apps uncovered 2,000+ high-impact vulnerabilities, 400 exposed secrets, and 175 instances of exposed personal data.
Common problems include broken access controls, exposed credentials, and cross-site scripting vulnerabilities. AI models have gotten better at generating functional code, but not necessarily safer code.
Takeaway: AI-generated software should undergo security review and testing. Use AI for prototyping or treat it as a coding assistant, not a replacement for secure development practices.
🚧 Companies Scale Back AI Support Agents
Data risks, hallucinations, and poor oversight are top concerns.
A recent Sinch report, covered by ITPro, polled over 2,500 customer service leaders and found that three in four companies have pulled back or abandoned AI agents in customer service.
The top reasons:
- Customer data exposure concerns (31%)
- Hallucinations and brand-reputation risks (22%)
- Lack of auditability and oversight (16%)
For example, Swedish fintech Klarna replaced its entire customer support team with an AI chatbot in 2024. Two years later, it is rehiring human agents. As Forbes reported, CEO Sebastian Siemiatkowski admitted that cost-cutting had pushed service quality too low.
Takeaway: Automate the routine. Protect the human touch where it counts.
📊 How to Hand Your Spreadsheets to AI
A real-world use case. Tools and workflows, ranked and tested.
By the Way Bakery, a gluten-free SME profiled in a recent WSJ article, replaced a wall of printed Excel spreadsheets with a custom AI planning tool. They used to update everything by hand.
Now, when an order arrives for 1,200 cakes, the system calculates ingredient requirements automatically. The company is candid about AI limitations: “AI still has hallucinations. You have to build guardrails in so that it’s not deciding to make 20,000 cakes on Monday.”
Total cost is a few hundred dollars a month in software fees, plus a few dollars in AI compute credits. That price point matters. A large food manufacturer might spend millions on a custom enterprise system. By the Way couldn’t afford that. AI closed the gap.
If you run your business on spreadsheets, your first major lift is removing the manual data input.
AI consultant Ruben Hassid tested Copilot, ChatGPT, Grok, and Claude for that job and documented his findings in Stop Learning Excel. Some tools were completely incapable. Others performed well. His top pick is Claude Cowork, via the desktop app on a Pro plan with Opus 4.7 and Adaptive Thinking selected.
Here is how to turn your data into a working spreadsheet:
- Open the Cowork tab and connect Google Drive so you can push the finished file directly to Google Sheets.
- Use this prompt structure: describe your purpose in one sentence, specify each sheet you need with its columns and calculations, add formatting instructions, then end with: “Before building, list your top 10 assumptions so I can sanity-check them, then execute.” That last step keeps you in control of the output.
Once your spreadsheet is ready, you can prompt Claude Cowork to manipulate the data, flag anomalies, or pull insights, without touching a formula yourself.
Thank you for reading today’s edition!
If this issue was valuable, pass it along to a fellow business owner. I’d love to hear your feedback at natalia@nataliabrattan.com.
See you next week,