Scaling Content - From 45 Pages to 1,000
Your site is growing and the commands that worked at 50 pages are producing warnings, overlaps, and costs you didn't expect at 200.
Content sites hit different walls at different sizes. A 50-page site has thin pages and no data. A 200-page site has pages fighting each other for the same keywords. A 500-page site has dead pages accumulating faster than anyone notices. The command that fixes one stage often doesn't exist yet at the previous one, and running it too early produces nothing useful. Where are you right now?
You ran `audit` and got a list of thin pages and underlinked pages. That's expected. GSC needs two to four weeks to populate, so commands like `data`, `enrich`, and `deduplicate` have nothing to work with yet. The only real lever at this stage is content: write longer pages, add internal links, create the pages that are obviously missing. But here's the complication: how do you know which pages to expand first when you have no ranking data at all?
GSC is live and you can see that two pages share several of the same keywords. One ranks position 14, the other ranks position 18. Neither is on page one. This is keyword cannibalization: both pages exist, so Google splits its attention between them instead of committing to either. Wire's `deduplicate` command classifies each overlap as a hard overlap (merge), soft overlap (differentiate), or minor overlap (report only). A hard merge archives the weaker page and creates a redirect. The combined page typically moves from position 14 to 7-9 within four to six weeks. Which situation fits yours?
Past 100 pages, a new problem appears: your content goes stale faster than you update it. A vendor page references a product version from eight months ago. A service page cites a statistic that has since changed. The `news` command gathers fresh information for a page, and `refine` integrates it. Together they cost about $0.36 per page. But running them on every page is wasteful. The question is which pages deserve fresh content and which ones can wait. At this stage, source concentration also appears: multiple pages citing only one or two sources, which makes the whole site fragile.
You started an `enrich` run on 200 pages, it processed 87, and then something interrupted it. Your laptop slept, the connection dropped, the process was killed. Without `--resume`, you start over and pay again for the 87 pages already done. The `--resume` flag reads a progress file at `.wire/progress-{command}-{topic}.json` and picks up from where it stopped. At this scale, another problem appears: comparison pages start stealing traffic from the vendor pages they're supposed to support. The "Acme vs Rossum" page outranks both Acme and Rossum individually.
At 500 to 1,000 pages, the audit stops being a diagnostic tool and becomes the workflow itself. You run it weekly. It tells you what to fix, and you run the specific commands it flags. Dead pages are the dominant problem: vendor companies shut down, products are discontinued, and pages that once ranked now return nothing. GSC also starts revealing keywords that have real search demand but no page on your site covers them. Those gaps are where new page creation pays off. The monthly cost at this scale is around $42 to $72, compared to $4,000 to $8,000 for a content editor.
The summary table maps each stage to its key command and the problem it solves: `create` at launch, `deduplicate` at first data, `news` and `refine` during growth, `enrich` and `newsweek` at scale, and `audit` as the workflow hub at maturity. The pattern is that you do not need all the commands on day one. Each command becomes relevant when the problem it solves actually appears. Running `deduplicate` before you have GSC data produces nothing. Running only `create` at 500 pages leaves cannibalization and staleness unaddressed.
Content sites follow a predictable growth curve. Each stage introduces problems the previous stage did not have. Wire's commands map directly to these stages. You use different commands as your site grows.
This guide follows two sites through their first year: AutoFix München (45 to 200 pages, local service business) and CompareStack (200 to 1,000 pages, SaaS comparison site). Both are fictional but based on real growth patterns.
Stage 1: Launch (1-50 Pages)
AutoFix launches with 45 pages: 15 service pages (bremsen, oelwechsel, inspektion), 12 location pages (schwabing, pasing, sendling), and 18 blog posts.
Problems at this stage:
- Pages are thin (200-400 words, written quickly to launch)
- No search data yet (GSC needs 2-4 weeks to populate)
- Internal linking is sparse (pages do not reference each other)
- No styleguide (Claude uses the built-in defaults)
What to run:
python -m wire.build # Build and verify structure
python -m wire.chief audit # Baseline health check
python -m wire.content create services/zahnriemen # Add missing service pages
Cost: ~$0.06 per page created. Creating 10 new pages costs ~$0.60.
What NOT to run: data, reword, deduplicate, enrich. These need GSC data that does not exist yet. Running them produces "no GSC data" warnings and wastes time.
AutoFix at Stage 1
The audit shows mostly structural issues:
HEALTH: services (15 pages)
- No GSC data
+ No dead pages
+ Titles OK
+ Descriptions OK
- 8 thin page(s) (< 200 words)
- 5 underlinked page(s)
The fix is content, not optimization. Write better pages. Expand thin ones. Add internal links manually or wait for Stage 2 when crosslink becomes useful.
Stage 2: First Data (50-100 Pages, Month 2-3)
Google has indexed AutoFix. GSC shows data for 30 of the 45 original pages. 15 pages get zero impressions.
Problems at this stage:
- First keyword overlaps appear (bremsen-reparatur and bremsen-wechsel share 5 keywords)
- Some pages rank on page 2 for valuable keywords
- Blog posts cannibalize service pages (a blog post about "Bremsen quietschen" outranks the service page)
- Location pages are thin and identical in structure
What to run:
python -m wire.chief data # Pull GSC metrics
python -m wire.chief audit services # Full SEO analysis
python -m wire.chief deduplicate services # Merge/differentiate overlapping pages
python -m wire.chief enrich services # Improve pages with GSC data
Cost: data and audit are free. deduplicate costs ~$0.12 per overlap pair. enrich costs ~$0.06 per page.
The First Overlap
The audit reveals:
ACTION: services
Merge (hard overlap, ratio > 0.4):
bremsen-reparatur + bremsen-wechsel
4 shared keywords, ratio 0.80, bremsen-reparatur gets 85%
This is classic keyword cannibalization. Two pages about brakes split Google's attention. Wire merges the weaker page into the stronger one, archives the donor, and creates a redirect. One strong page replaces two weak ones.
Before merge: bremsen-reparatur ranks position 14, bremsen-wechsel ranks position 18. Neither page 1.
After merge: the combined page has more content, more keywords, and no internal competition. Typical result: position moves from 14 to 7-9 within 4-6 weeks.
Stage 3: Growth (100-200 Pages, Month 4-6)
AutoFix expanded from services into blog content. CompareStack has reached 200 vendor profiles.
Problems at this stage:
- Cross-topic cannibalization (blog posts compete with service pages)
- News staleness (vendor pages reference year-old product updates)
- Source concentration (5 vendor pages cite only ADAC)
- Content gaps appear (GSC shows demand for "TÜV Vorbereitung München" but no page exists)
What to run:
python -m wire.chief audit # Cross-topic audit (no topic arg = all topics)
python -m wire.chief news services # Gather fresh news for service pages
python -m wire.chief refine services # Integrate news into pages
python -m wire.chief crosslink services # Add internal links to underlinked pages
python -m wire.chief sanitize # Fix any broken links
Cost: news costs ~$0.30 per page (web search + junior evaluation + senior synthesis). refine costs ~$0.06 per page. crosslink costs ~$0.06 per underlinked page.
AutoFix at 150 Pages
Monthly costs stabilize:
| Command | Pages affected | Cost per page | Monthly total |
|---|---|---|---|
| enrich | ~30 (top 20%) | $0.06 | $1.80 |
| news + refine | ~15 (services only) | $0.36 | $5.40 |
| reword | ~20 (top 50% with GSC data) | $0.06 | $1.20 |
| crosslink | ~5 (newly underlinked) | $0.06 | $0.30 |
| Total | ~$8.70/month |
Data, audit, and sanitize are free. The monthly cost for a 150-page site is under $10.
CompareStack at 200 Pages
CompareStack hits different problems at the same page count. With 200 vendor profiles, keyword overlap is everywhere:
HEALTH: vendors (200 pages)
+ GSC data loaded (12,400 keywords)
- Cannibalization: 47 overlap pairs (3+ shared keywords)
- 12 dead page(s) below threshold
- 8 pages with source concentration
47 overlap pairs means 47 pairs of pages fighting each other for the same keywords. Wire classifies each:
- 8 hard overlaps (merge the weaker page)
- 23 soft overlaps (differentiate both pages)
- 16 minor overlaps (report only, no action)
Running deduplicate resolves the 31 actionable pairs: 8 merges ($0.12 each = $0.96) and 23 differentiations ($0.12 each = $2.76). Total: $3.72 for a one-time cleanup that would take a human editor weeks.
Stage 4: Scale (200-500 Pages, Month 7-9)
CompareStack adds comparisons. Every vendor pair generates a comparison page. 200 vendors produce up to 19,900 possible comparisons, but only ~80 have search demand.
Problems at this stage:
- Comparison pages cannibalize vendor pages (the "Acme vs Rossum" comparison steals traffic from both vendor pages)
- The batch pipeline matters now (running commands one page at a time is too slow)
- Resume becomes essential (a 200-page enrich run takes 2 hours; interruptions happen)
- Newsweek reports become valuable (weekly market intelligence across all vendors)
What to run:
python -m wire.chief enrich vendors --resume # Resume interrupted run
python -m wire.chief newsweek --from 2026-03-03 --to 2026-03-10
python -m wire.chief reword vendors # Tiered: top 20% full SEO, next 30% light
The Resume Pattern
At scale, batch commands run for hours. The --resume flag tracks progress in .wire/progress-{command}-{topic}.json:
# Start enrich on 200 vendor pages
python -m wire.chief enrich vendors
# ... processes 87 pages, then your laptop sleeps
# Resume from page 88
python -m wire.chief enrich vendors --resume
Progress files are cleaned up on successful completion. If a run crashes, delete the progress file to start fresh, or use --resume to continue.
Newsweek Reports
With 200+ vendor pages, Wire has enough news to produce weekly market intelligence reports:
python -m wire.chief newsweek --from 2026-03-03 --to 2026-03-10
This runs a three-phase pipeline: extract news from all topics, synthesize into a thematic report, and optionally review for quality. Cost: ~$3 per full report. Output: a publication-ready markdown page at docs/news/2026-03-10-news.md.
Stage 5: Maturity (500-1,000 Pages, Month 10-12)
CompareStack at 800 pages. AutoFix at 200 pages with strong local rankings.
Problems at this stage:
- Dead pages accumulate (vendor companies shut down, products are discontinued)
- The audit becomes the primary workflow tool (not individual commands)
- Content gaps drive new page creation (GSC reveals keywords nobody owns)
- Cost optimization matters (not every page deserves a full enrich pass)
The Monthly Workflow at Scale:
# Week 1: Data + Audit
python -m wire.chief data --force
python -m wire.chief audit
# Week 2: Fix problems the audit found
python -m wire.chief deduplicate
python -m wire.chief sanitize
# Week 3: Improve content
python -m wire.chief enrich vendors --resume
python -m wire.chief enrich guides --resume
# Week 4: News + Report
python -m wire.chief news vendors
python -m wire.chief refine vendors
python -m wire.chief newsweek --from ... --to ...
Cost at Scale
| Site size | Monthly enrich | Monthly news | Monthly reword | Newsweek | Total |
|---|---|---|---|---|---|
| 50 pages | $1.50 | $5.40 | $0.60 | - | ~$7.50 |
| 200 pages | $6.00 | $10.80 | $2.40 | $3.00 | ~$22.20 |
| 500 pages | $15.00 | $18.00 | $6.00 | $3.00 | ~$42.00 |
| 1,000 pages | $30.00 | $27.00 | $12.00 | $3.00 | ~$72.00 |
Compare to manual alternatives: a content editor costs $4,000-8,000/month. A junior SEO analyst costs $3,000-5,000/month. Wire at 1,000 pages costs $72/month and never calls in sick.
The real cost is the Anthropic API key. Wire's analysis, audit, build, and lint commands are all free.
What Changes at Each Stage
| Stage | Pages | Key command | Problem solved |
|---|---|---|---|
| Launch | 1-50 | create |
Generate initial content |
| First data | 50-100 | deduplicate |
First keyword overlaps |
| Growth | 100-200 | news + refine |
Content freshness |
| Scale | 200-500 | enrich + newsweek |
Batch improvement + intelligence |
| Maturity | 500-1,000 | audit (as workflow hub) |
Dead pages, gaps, optimization |
Wire's commands are designed for this progression. You do not need all of them on Day 1. Use what your current stage requires and ignore the rest until the problems appear.