BT · Amazon
200,000 — Lines of code written
Independently reportedTelecom · Code generation · LLM · In production
Code generation
Finger Lakes Daily News · TOTIB
Vendor-reportedMedia & entertainmentIn productionLLMLive 2026
Vendor-reported. The customer is named and the numbers are quoted from the source page, but the account comes from the vendor. No independent confirmation.
Reported by totib.com (vendor self-report). Checked against the source page on 2026-09-05. 16 of 18 figures below appear on that page word for word; the rest are marked.
The problem
The news site had loyal readership but virtually no technical SEO infrastructure, structured data, performance optimization, or monitoring.
What was deployed
- Vendor
- TOTIB
- Models
- Claude
- Technique
- LLM
- Build or buy
- Bought off the shelf
- Deployment
- Custom build
- Data used
- live WordPress data, GA4, Google Search Console, BigQuery warehouse, NWS and Open-Meteo
- Scale
- Local news site covering the Finger Lakes, NY region; 387K+ requests/day logged to BigQuery
- Timeline
- 3 months
What changed
Each row is quoted from the source. Figures we could not find on the page in those words are marked — they are kept, not deleted, so you can judge them.
99Mobile PageSpeedfrom 54
“54→99 Mobile PageSpeed +45 points”
Quoted word for word from the source
7.4%Organic CTRfrom 4.6%
“7.4% Organic CTR Up from 4.6%”
Quoted word for word from the source
1.6MStorm Day Requestssingle day
“1.6M Storm Day Requests”
Quoted word for word from the source
87.5%cache hit rate
“87.5% cache hit rate”
Quoted word for word from the source
12+Schema Types implemented
“12+ Schema Types Full structured data”
Quoted word for word from the source
under 10msTotal Blocking Time (TBT)from 730ms
“GTM replaced with Zaraz, cutting 730ms of blocking time. TBT dropped to under 10ms.”
Not found on the source page in these words
1.2sObit archive LCPfrom 5.6s
“Obit archive LCP: 5.6s → 1.2s”
Quoted word for word from the source
1.4sWeather page LCPfrom 6.5s
“Weather LCP: 6.5s → 1.4s”
Quoted word for word from the source
0sMain-thread scriptingfrom 3.8s
“Main-thread scripting: 3.8s → 0s”
Quoted word for word from the source
~120Origin requests for now-playing APIfrom 83,000/day
“Origin requests dropped 99.8%—from 83,000/day to ~120.”
Quoted word for word from the source
2-5msNowplaying Edge API response timefrom 660ms
“Response: 660ms → 2-5ms”
Quoted word for word from the source
<60 secondsStorm article auto-generation time
“Storm articles in <60 seconds”
Quoted word for word from the source
82%Share of Voice for Crime & Police
“82% Crime & Police Dominant in local crime search”
Quoted word for word from the source
9,602Peak Storm Day Clickspeak storm day
“9,602 Peak Storm Day Clicks”
Quoted word for word from the source
176News Tab clicksApr 1
“FLDN's News tab performance hit its all-time high: 176 clicks from 12,961 impressions, driven by auto-published weather coverage that Google surfaced within minutes.”
Quoted word for word from the source
8.9Average Position (GSC)from 13.2
“13.2 → 8.9 (33% better)”
Quoted word for word from the source
10.6%Police Blotter CTRfrom 8.3%
“Police Blotter CTR Crime reporting CTR surged as structured data and focused indexing reinforced topical authority. 8.3% → 10.6%”
Not found on the source page in these words
over 60 sessions/weekReferral traffic from ChatGPTfrom ~3 sessions/week
“referral traffic from ChatGPT has grown from ~3 sessions/week to over 60—a 20x increase.”
Quoted word for word from the source
Difficulties and limits
- Lighthouse scores hovered around 54 on mobile and 69 on desktop, with no JSON-LD structured data, no canonical URL strategy, and no mechanism to prevent thin or duplicate content
- Entertainment and national news sections were syndicated RSS feeds (identical content on hundreds of sites), generating 683,000 impressions/month at 0.3% CTR from position 28 - diluting topical authority and wasting crawl budget
- Google typically takes 3–6 months to fully reflect structural changes in rankings.
Sources
- totib.com — vendor self-report
Similar deployments
allpay · Microsoft
increased by 10% — developer productivity
Vendor-reportedBanking & finance · Code generation · LLM · In production
iTromsø · IBM, VC2
10 minutes per day — time spent reviewing potential stories
Independently reportedMedia & entertainment · Document processing · NLP · In production
Hearst Newspapers · OpenAI
more than 1,000 — requests handled
Independently reportedMedia & entertainment · Marketing content · LLM · In production
Trend Assist · Datavise
95 — User Satisfaction Rate
Vendor-reportedMedia & entertainment · Customer support · LLM · In production
Roboto Games · nunu.ai
Vendor-reportedMedia & entertainment · Process automation · AI agents · Scaled