Case study
From GEO 44 to 87 in 11 days: how eb-becker.de got recommended by AI
Published May 13, 2026 · by Gerrit Halfmann
eb-becker.de, a nutrition consultancy in Schermbeck, Germany, went from being recommended by 0 of 7 AI platforms to 6 of 7, and from a GEO Score of 44 to 87, in 11 days. Three concrete fixes did the work. The full re-scan history is public.
GEO Score
44 → 87
+43 absolute · +98%
AI platforms
0 → 6
out of 7 at the time
Time invested
~3h
over one afternoon
The starting point
On May 2, 2026, eb-becker.de scored 44/100 on AskMention’s GEO scan. The site was a clean, modern WordPress install with reasonable SEO. It ranked decently on Google for local German nutrition queries.
But when we asked ChatGPT, Gemini, Perplexity, and four other AI platforms “Recommend a good Ernährungsberatung in Schermbeck” — eb-becker.de was mentioned by zero of them. Every AI recommended other businesses instead: Frank Pudel, Nadia Abdereman-Tange, Praxis Sauer, Naturheilpraxis Schleifer, and others.
That’s the gap: ranking #1 on Google doesn’t mean AI knows you exist. The signals are different.
The three fixes
AskMention’s action plan flagged dozens of items, but three were marked high-impact. We implemented only those:
1. Consolidate the homepage to a single H1
The site had multiple H1 tags. AI crawlers use heading hierarchy to understand what a page is about — multiple H1s dilute the entity signal. We collapsed them to one clear H1 stating the business and service.
2. Expand JSON-LD sameAs for Knowledge Graph entity resolution
The schema markup had an Organization block but only one external link. We added sameAs links to LinkedIn, Google Business, the founder’s social profiles, and the regional health association — telling AI “this Organization is the same entity referenced on those other authoritative pages.”
3. Optimize meta descriptions (120–158 chars, keyword + location + brand)
Meta descriptions on key pages were either missing or too generic. We rewrote them to 120–158 characters with the service category, the location (Schermbeck), and the business name — the three pieces AI uses to decide who to recommend for “X in Y” queries.
What happened
We re-scanned the same evening. Score: 86/100. From 0 of 7 platforms to 6 of 7 in a single afternoon.
The journey from 86 wasn’t linear — over the following week the score fluctuated between 65 and 86 as AI platforms re-indexed and re-evaluated. By May 13 it settled at 87/100, with 6 of 7 AI platforms recommending eb-becker.de when asked about local nutrition consultancies.
Full scan history
Every dot is a real re-scan in our system.
| Date | GEO Score | Note |
|---|---|---|
| May 2, morning | 44 | Baseline before fixes |
| May 2, afternoon | 86 | After 3 fixes implemented |
| May 3–7 | 86 | Stable |
| May 8 | 72 | AI re-indexing fluctuation |
| May 9 | 65 | Continued fluctuation |
| May 13 | 87 | Settled higher than initial peak |
What this tells you
- The most impactful fixes are technical and dull. Single H1, sameAs schema, meta descriptions. No content rewrite, no link-building.
- AI re-indexing isn’t instant or monotonic. The score dipped before it stabilized higher. This is normal — AI platforms re-evaluate over days, not seconds.
- The same three fixes don’t work everywhere. They worked here because eb-becker had an entity-resolution problem (AI didn’t know who they were). For sites with content-quality problems or zero off-site authority, the fixes are different.
- You can verify this case study yourself. The public scan URL is live and the score is recomputed every time we re-scan. No before/after marketing trickery.
Want this for your site?
Run a free scan. If you score below 70/100, you have specific fixes to make. We’ll show you which ones, ranked by impact.
Scan my site → free