Claude
Rebuilt the site as static HTML and runs the scheduled automations: publishing, LinkedIn, Business Profile posts and PR triage.
D-05 · B2B consultancy automation stack case study
39 bytesto107 KB
A B2B consultancy whose website was invisible to AI search, and whose publishing, social and outreach all depended on one person doing everything by hand.
Live, running weeklyB2B consultancy · AI search & automationUS & PakistanAug 2026 onwards
AI search readinessWorkflow automationContent operationsAnalyticsTechnical SEO
Platforms
01
At a glance
Aug 2026 onwards107 KBHTML crawlers can read, up from 39 bytes29 Aug 2026
9AI crawlers explicitly allowedSep 2026
5 / wkLinkedIn posts from a self-replenishing queueSep 2026
11Steps in every automated publishing runSep 2026
The short answer
The site was rebuilt from a JavaScript-only builder into static HTML, taking what crawlers could read from 39 bytes to 107 KB and 2,071 words, with nine AI crawlers explicitly allowed. It was then connected to a set of governed automations: a twice-weekly publishing run that writes, builds, publishes, verifies and requests indexing; a LinkedIn system posting five times a week from a self-replenishing queue; Business Profile posting; branded image generation; and a journalist-request triage flow that drafts replies but never sends them.
02
Where things stood
the starting point- Invisible to AI search: the builder served an empty shell, so crawlers that do not run JavaScript saw 39 bytes.
- Broken measurement: the analytics tag reported to a property nobody could open, and there was no Search Console property.
- Everything by hand: posts, social, Business Profile updates and press requests all waited for one person.
03
What changed
results107 KB HTML crawlers can read, up from 39 bytes
29 Aug 2026
- Every word now in the source HTML: 107 KB and 2,071 words readable by every crawler, from 39 bytes.
- Nine AI crawlers explicitly allowed, including the ChatGPT, Claude, Perplexity and Google AI agents.
- Twice-weekly articles published automatically: written to a house style, built in the site's design, schema-checked, published, verified and submitted for indexing on the day.
- LinkedIn on autopilot: five posts a week across four named series, each with a branded image, from a queue that refills itself.
- Branded images in three formats generated for every post: feed, Business Profile and article hero.
- Every new article also posted to Google Business Profile with a tracked link.
- Journalist-request triage: hourly weekday scans of five PR platforms, with draft replies waiting for approval, never sent automatically.
- Measurement restored: a working GA4 property and a DNS-verified Search Console domain property.
04
In numbers
hover or tap for exact valuesWhat crawlers could read
Before and after the static rebuild, Aug 2026
HTML returned to crawlers (KB)0.039 → 107
Before
After
Crawlable words on the homepage0 → 2,071
Before
After
View as table
| Measure | Before | After |
|---|---|---|
| HTML returned to crawlers (KB) | 0.039 | 107 |
| Crawlable words on the homepage | 0 | 2,071 |
Automated output per week
Scheduled runs, Sep 2026
LinkedIn posts5
Articles published and indexed2
Business Profile posts2
View as table
| Item | Value |
|---|---|
| LinkedIn posts | 5 |
| Articles published and indexed | 2 |
| Business Profile posts | 2 |
The stack
Sep 2026
9AI crawlers explicitly allowed
3Image formats rendered per post
5PR platforms triaged hourly on weekdays
4Governance rules on every automation
Want a result like this for your business?107 KB HTML crawlers can read, up from 39 bytes. Book a free 15-minute diagnostic and we will tell you what would move your number.
05
How we did it
diagnose, model, build, compound01Diagnose
A crawl of what search engines and AI assistants actually receive, an analytics audit, and a map of every repeated weekly task.
02Model
One rule for every automation: a named owner, a human approval gate on anything that leaves the business, an audit log and a kill switch.
03Build
A static rebuild with one canonical host and 301s, schema on every page, an AI-crawler policy, a rebuilt GA4 property and DNS-verified Search Console; then the publishing run, the LinkedIn system, the image renderer, Business Profile posting and PR triage.
04Compound
Runs on a schedule in the cloud, verifies its own output after every publish, requests indexing, and reports each run.
06
How we leveraged the platforms
what each one did in this engagementHostinger
Hosting: every published page is verified live and the cache cleared after each run.
Google Workspace
The inbox the PR-triage flow reads and drafts replies into, for human approval.
Google Analytics 4
A rebuilt property, so traffic from every channel is finally visible.
Google Search Console
DNS-verified domain property; each new article is submitted for indexing the day it goes live.
Google Business Profile
Every new article is posted to the profile with a tracked link.
Five posts a week across four named series, each with a branded image.
07
Who we worked with
by level and roleFounder and managing director
What happens nextMove deployment to Git so no browser is needed in the loop, add IndexNow, and push enquiries into a CRM with a lead event in analytics.
08
Keep exploring
the services, proof and reading behind this resultServices behind this result
Applied AI & AutomationAssistants and workflow automation with approval gates and audit logs.Digital Product DevelopmentWebsites, apps and platforms built fast, crawlable and accessible.09
Common questions
Why rebuild instead of adding SEO plugins?
Because the problem was the delivery, not the content. If crawlers receive an empty page, no plugin can fix it. Static HTML makes every word readable by Google and by AI crawlers that do not run JavaScript.
Do the automations post without anyone checking?
Only where the owner chose that. Anything that speaks to a third party, such as press replies, stops at a draft for human approval. Every run is logged and can be switched off.
What does an automated publishing run do?
It picks the due article, edits it to the house style, builds it in the site's design with schema, publishes it, clears the cache, verifies the live page, requests indexing, posts it to the Business Profile and shares it on LinkedIn.
Can this be built for another business?
Yes. It is the same pattern we use for clients: fix what machines can read first, then automate the repeated work with approval gates.
Next step