Dead lines
47%
146 of 310 posts had not earned a single organic click in twelve months. Most were reactions to that week's news, stale within a month.
Fault 1 of 4
A fictional business. The figures are a modelled scenario, not a client result.
Gridlark Analytics · Data-pipeline analytics SaaS, 120 staff · Austin
+186% organic trial sign-ups, named in 41% of AI answers
18 months · North America · Content Engine plan · Topic clusters · Content pruning · AI-search visibility · Original research
§ 1The trouble on the line
Its blog chased trends with no topic structure, and when engineers asked AI assistants how to monitor pipelines, the answers named competitors.
Gridlark sells pipeline observability to data teams, and its buyers research for weeks before they book a demo. The blog published three times a week, but on whatever was trending: conference recaps, hot takes on the latest AI launch, listicles. Nothing tied one post to the next, so search engines never saw Gridlark as an authority on anything. Then engineers started asking AI assistants instead of searching, and the assistants, quoting competitors' how-to guides, put them straight through to someone else.
§ 2Fault report
47%
146 of 310 posts had not earned a single organic click in twelve months. Most were reactions to that week's news, stale within a month.
Fault 1 of 4
38 pairs
38 pairs of posts competed for the same query. Nine different posts targeted “data pipeline monitoring”, and not one of them reached page one.
Fault 2 of 4
0 hubs
No pillar pages and no internal-linking plan. Posts linked to the homepage and to pricing, almost never to each other, so no topic ever built up authority.
Fault 3 of 4
6%
Across 200 prompts engineers really ask, AI assistants named Gridlark in 6% of answers. They cited competitors' step-by-step guides because Gridlark had none to cite.
Fault 4 of 4
§ 3Operating procedure
Built on our Content Engine plan: the content and AI-search engagement for companies whose buyers research long before they talk to sales.
3.1
Six hub pages, one for each job an engineer needs done: monitoring, data quality, lineage, alerting, warehouse cost and schema changes. Around them, 84 spoke pages that each answer one specific question in full, mapped from 1,900 queries, support tickets and sales-call notes. Every spoke links up to its hub, and every hub links down to all its spokes.
3.2
142 posts were merged into stronger pages or retired with a 301 to the closest spoke. The 31 best were rewritten as spokes, and the nine competing “pipeline monitoring” posts became one page that finally ranked.
3.3
Every spoke opens with a two-sentence direct answer, then steps, code and a comparison table: the shape AI assistants quote. Named authors with real bylines, TechArticle and FAQPage schema, and a monthly check of 200 prompts across four assistants.
3.4
A yearly Pipeline Reliability Report, built from anonymised incident data across 900 pipelines. Its figures are the ones other writers and AI answers now quote, and every citation leads back to a hub.
§ 4The switchboard
On the left, twelve calls engineers place every day. On the right, six hub jacks. Choose a call, then the hub that should answer it, and the cord connects it to that hub's spoke page. Every call left unplugged rings out to a competitor.
What an AI assistant answers
Put a call on the line to hear the answer.
Esc clears the line.
Choose a call to put it on the line.
§ 5Shift log
Months 1–2
Crawl of all 310 posts, a year of Search Console data, a 200-prompt AI baseline and 1,900 questions pulled from search, support and sales calls.
Months 3–4
Six hubs and 84 spokes mapped, cannibalising pairs resolved, the merge-and-redirect plan signed off with the product team.
Months 5–8
Hubs and the first 40 spokes live, 142 old posts merged or retired, internal links rebuilt from hub to spoke and back.
Months 9–13
The remaining 44 spokes, the first Pipeline Reliability Report, author pages and schema across the cluster.
Months 14–18
Monthly prompt checks, refreshes for the 20 spokes that slipped, the second report and a webinar series built on the hubs.
§ 6Results
Organic trial sign-ups / month
+186%
210 → 600 a month
AI-assistant mention rate (200 prompts)
6% → 41%
Share of 200 tracked prompts, across 4 AI assistants
Non-branded organic clicks / month
9,800 → 31,500
Per month, quarter by quarter
Topic clusters wired
6 hubs · 84 spokes
31 rewritten from old posts, 53 new
Modelled scenario for a fictional business — not a client result. Rankings, traffic and AI-assistant mentions are never guaranteed; see our Terms.
| Prompt, as typed | Hub | AI answer · month 1 | AI answer · month 18 | Google rank |
|---|---|---|---|---|
| how to monitor airflow dag failures | Pipeline monitoring | not named | #1 of 3 | #24 → #2 |
| data pipeline monitoring tools | Pipeline monitoring | #4 of 4 | #1 of 4 | #17 → #3 |
| data quality tests in dbt | Data quality | not named | #2 of 3 | #31 → #4 |
| column level lineage explained | Data lineage | not named | #1 of 2 | #22 → #1 |
| reduce alert fatigue data team | Alerting & on-call | not named | #1 of 3 | #40 → #3 |
| why is my warehouse bill so high | Warehouse cost | not named | #2 of 4 | #28 → #5 |
| handle schema drift in streaming | Schema changes | #3 of 3 | #1 of 3 | #15 → #2 |
| upstream schema change broke my pipeline | Schema changes | not named | #1 of 2 | #36 → #4 |
“#1 of 3” means Gridlark was the first of three tools the answer recommended, checked across four AI assistants. Google rank is the matching spoke page's position for the same query, month 1 → month 18.
Is your blog busy but your pipeline quiet? Start with the Content Engine plan, or tell us which questions your buyers ask before they call.