Release Notes

What we shipped

• Country-scoped keyword data. Search Console pulls, keyword filters, and competitor analysis are now scoped to the site's target country. A Bali hotel no longer sees noise from people searching in random other markets.
• Live SERP verification. The keyword positions shown in the report are checked against a real Google search from the customer's primary country, not just a 28-day GSC average that blends locations.
• Daily-or-weekly email frequency. Customers can now pick how often they want their recap email. Same content, customer's cadence. Setting lives in site settings, switchable any time.
• Auto-translated recap emails. If the site is in Spanish, Portuguese, French, Hebrew, etc., the email arrives in that language. Picked up from the site - nothing to configure.

What it means for customers

The rank numbers in the report now match what a real searcher in the customer's country would see, instead of a blended average that hides what's happening in the market they actually care about. Customers who want a daily nudge pick daily. Customers who want a clean weekly roll-up pick weekly. And the recap arrives in the site's own language, so a Brazilian customer isn't reading their progress report in English.

What we shipped

• The stats section no longer says "down from X last week" when impressions or clicks dip. Week-to-week swings on a small site are mostly noise, and showing them as losses is misleading. We now show the absolute number instead.
• Keyword wins are gated by a real signal floor. A keyword has to clear at least 5 impressions in the week to count as movement. One or two stray impressions can no longer surface as a "ranking improvement."
• Wins are also geo-filtered. A North-based business will no longer see a Jerusalem-only keyword bubble up as a win because somebody on holiday clicked it once.

What it means for customers

The weekly recap stops creating drama where there isn't any. No "you dropped 12%" when nothing actually changed, no "you're ranking for X" when X is a one-click ghost. What's in the report matches what's actually happening on the site. That's the only way it stays useful three months from now.

LinkedIn post (ready to copy)

Most SEO reports have a quiet design problem: they manufacture emotion.

A 6% impression dip on a small site is statistical noise. But the report says "down 6% from last week" and the founder spends Monday morning anxious. A keyword catches one stray click from someone in the wrong country, and the report says "you're now ranking for [keyword]" and the founder spends Monday morning excited about a number that will be gone next week.

Neither of those moments is real. Both of them erode trust over time, because eventually the founder figures it out.

This week at Scaup we shipped the unglamorous fix:

- We don't frame week-over-week as "down" when it's noise. Absolute number only.
- Keyword wins need a real signal floor before they show up. Single-impression flukes are filtered out.
- Wins are matched to the customer's country. A North-based business doesn't see a win from a city they don't serve.

The recap should match reality. If a real thing improved, say so. If nothing meaningful happened, say nothing happened. The reason people stop opening dashboards isn't that they're bored. It's that they stopped trusting the numbers.

What we shipped

• A substance gate. Before drafting a new page, the system checks for a real user question and concrete facts to use. If the raw material is too thin, we skip the page instead of publishing filler.
• A hard cap on related variations. We never produce more than a handful of pages per topic, and each has to use materially different facts (no swapping a city name and calling it a new page).
• Anti-AI-writing filters extended. Em dashes, "leverage," "seamless," rule-of-three sentences, and "ing-analysis" patterns are stripped before anything reaches a customer site.
• Every prompt now opens with a one-line standard: every page must be defensible under Google's helpful-content guidance.

What it means for customers

The most common worry we hear from founders building with AI tools is that Google will flag their AI-written pages and tank their rankings. The honest answer is that Google does not flag AI content for being AI. It flags content that doesn't help users. These changes make it structurally impossible for Scaup to produce the patterns Google does penalize, so customers can run the system without worrying about a policy hit.

Deep-dive article

LinkedIn post (ready to copy)

One of the most common questions I get from founders building with AI tools: "Will Google flag my site as AI content and tank my rankings?"

It's a fair worry. Here's what's actually true.

Google's policy (restated in their March 2024 spam update) doesn't target AI. It targets "scaled content abuse." Mass-produced thin pages. Near-duplicates with a swapped city or product name. Content that doesn't answer the search. Whether AI wrote it, you wrote it, or both, the rules are the same.

What actually hurts your rankings isn't who wrote the page. It's whether the page is useful.

That's the bar we hold ourselves to at Scaup. When the raw material isn't there to make a page genuinely useful for a real user, we publish nothing that day instead of filling space. That's harder to build than "always publish something," but it's the only honest way to run automated SEO.

The point isn't to avoid AI. The point is to never produce the pattern that gets sites demoted, by design.

Wrote up the full story (what Google actually penalizes, what to check on your own site, how we think about this): https://scaup.io/does-google-flag-ai-content