82%
of 140 posts were product news
115 of 140 posts were release notes, webinar invitations and hiring news. The 25 how-to posts shared no structure, had no code that could be copied and averaged 1.4 internal links each.
Specimen 1 of 4
A fictional business. The figures are a modelled scenario, not a client result.
FernlogLog analytics SaaS, 60 staff · Tallinn
+261% trial sign-ups from organic search
18 months · Europe · Content Engine plan
Fig. 1 — The crozier unrolls: one stem, four topic hubs, 20 subtopic pages.
01the challenge
Developers asking how to parse or alert on logs found competitors' docs and Stack Overflow. Fernlog's blog was product news, never cited by AI assistants.
Every week, thousands of engineers search for how to parse a stack trace, tame a noisy alert or cut a log bill. Fernlog's product does all three, but its blog was a feed of release notes and webinar invitations. Search engines sent those developers to competitors' docs and Stack Overflow threads, and when they asked an AI assistant instead, the answer cited everyone but Fernlog.
parse multiline stack traces
AI answers to 120 developer prompts · sources cited, before
An answer can cite several sources.
02diagnosis
82%
of 140 posts were product news
115 of 140 posts were release notes, webinar invitations and hiring news. The 25 how-to posts shared no structure, had no code that could be copied and averaged 1.4 internal links each.
Specimen 1 of 4
0 / 120
AI answers citing Fernlog
Each month we ran 120 real developer prompts through four AI assistants. Not one answer cited Fernlog; competitors' docs appeared in 71.
Specimen 2 of 4
61%
of docs URLs not indexed
The docs lived on a client-rendered subdomain: 61% of its URLs were not indexed, and code samples on the blog were screenshots that neither search engines nor AI crawlers could read.
Specimen 3 of 4
1,900
developer questions, almost none answered
Support tickets, community threads and sales calls held 1,900 distinct developer questions. They clustered into four themes — parsing, alerting, retention and cost — and Fernlog had a page for almost none of them.
Specimen 4 of 4
03the strategy
Built on our Content Engine plan: topical content and AI-search work for companies whose buyers research on their own before they talk to sales.
from service:"checkout" level:error| parse json(message) as m| stats count() by m.error_code| alert when count > 50 within 5mFour hubs mapped from the 1,900 questions: parsing, alerting, retention and cost. Each hub page answers the broad question and links down to subtopic guides, which link back up and across, so authority flows along the frond instead of pooling in the newsroom.
Every guide opens with a working answer — a query, a config block or a CLI command, in real text that can be copied — then explains why it works. Samples are tested against the product in CI, so they never go stale.
Answer-first paragraphs, one claim per heading, dated benchmarks, and TechArticle and HowTo schema. We tracked the same 120 prompts every month and rewrote pages that were retrieved but not cited.
An anonymised benchmark of ingestion volume per service, built from aggregate platform data, became the most-cited page on the site: the cost-per-GB figures other blogs and AI answers now quote.
04the topic map
Each hub is a frond and each subtopic page a leaflet. Switch between the start and month 18 to see which pages rank in Google and which AI-search answers cite Fernlog.
Key
Topic hubs
Turning raw lines into fields you can query.
JSON logs
AI answers to 6 prompts on this subtopic
05the timeline
Months 1–2
1,900 questions pulled from support tickets, community threads and sales-call notes, clustered into four hubs and twenty subtopics. Baseline: 120 AI prompts and 380 Google queries.
Months 3–5
A hub page and six guides, each opening with a runnable answer. The docs moved to server rendering, and code samples became text tested in CI.
Months 6–8
Five guides plus a library of copy-paste alert rules. The first AI citations arrive, for the stack-trace and JSON guides.
Months 9–12
Nine more guides and the cost-per-GB benchmark, built from anonymised platform data and launched with a methodology page.
Months 13–18
Monthly prompt runs: pages retrieved but not cited were rewritten answer-first, stale samples were refreshed, and release notes moved to a changelog.
06results
Trial sign-ups from organic search
+261%
64 → 231 trial sign-ups a month
AI answers citing Fernlog (of 120)
0 → 41
The same 120 prompts across four AI assistants, re-run every month.
Developer queries in Google's top 10
9 → 146
Of 380 tracked developer queries.
Organic sessions a month
11,800 → 52,400
Organic sessions a month to guides, hubs and docs.
Modelled scenario for a fictional business — not a client result. Rankings, AI citations and traffic are never guaranteed; see our Terms.
StartMonth 18
Our Content Engine plan builds topic hubs and AI-search visibility for technical products, from the first question map to the monthly citation report.