First: what did I actually do with these 6 so far?

Nothing built yet — I only marked them as adopted. What already shipped on the site (trust strip, "what moves the price", reviews on money pages, mobile call bar) came from my own earlier plan, not from this doc. These 6 go in the deep-deck pilot (1 New Britain page, preview first, you approve, then it spreads or it dies). This page exists so you can approve with eyes open.

The 3 spam rules Google actually punishes — kept in plain words

Module 1 — Day-by-day timeline ("Day 1: demo → Day 6: grout + walkthrough") SPAM RISK: LOW

Buys: the single most shared, screenshotted, time-on-page format in contractor content. Kills the biggest buyer anxiety ("how long will my bathroom be torn up?").

Search value: catches "how long does X take" questions — real query family.

⚠ Alert: the timeline is per-service, not per-town — so all 9 bathroom pages would share ~90% of it. That's acceptable (it's short, structured, and clearly the same company's process). I will NOT invent a fake town timeline per page just to look unique — that's thin-content thinking in reverse.

Module 2 — "What's in your estimate" systems table (KERDI-BOARD, ANSI A118.10, DCOF pan tile, uncoupling membrane) SPAM RISK: LOW

Buys: reply-proof against "you're 20% more than the handyman" — the table shows what the money actually is. Also fills 6 of the 11 missing trade-words from my entity sheet in one module.

Search value: exact trade terms = entity signals Google can attach us to.

⚠ Alerts: (a) only list systems we actually use on that service — 12 service-specific versions, not one big table everywhere; (b) Schluter by name is allowed as system naming, but no "Schluter-certified installer" style credential claims (banned, because untrue); (c) keep it a table, not a pdf-dump of standards.

Module 3 — Symptom checker ("stain on the ceiling below = the pan, not the tile") SPAM RISK: MED

Buys: captures homeowners in "do I need a pro?" mode and converts panic into a call. Very high formatting value (checkbox list).

Search value: informational queries graders research before hiring ("cracked grout corners shower", "loose tiles floor hollow").

⚠ Alerts: (a) intent mismatch risk — a repair-question visitor on a sales page bounces if the page doesn't welcome them; the module ends with the honest soft ask ("text/send 3 photos, we'll tell you straight") rather than a hard sell; (b) symptom sets must differ per service (shower ≠ floor ≠ grout) — 12 versions, not one; (c) only claim the failure types that actually match the town's house age (e.g., mortar-bed pans in pre-1960 stock) — no diagnosing houses we've never seen.

Module 4 — Jump menu (Cost · Timeline · Photos · FAQs chips) SPAM RISK: ~ZERO

Buys: long pages become scannable; mobile users stop hunting. Pure UX — like a table of contents in a book.

Search value: minor bonus (can earn sitelinks/anchor understanding). Not a ranking lever by itself, and I won't pretend it is.

⚠ Alert: none, realistically. Identical nav on every page is fine — it's interface furniture, not content.

Module 5 — The 6-question town profile (house age, main problem found, building type, permit office, what locals choose, drive time) SPAM RISK: LOW — this is the anti-doorway engine

Buys: this is what makes every town page provably different from every other: different era → different substrate → different failure → different job. It's also where your quote expectations get set correctly by town.

Search value: the strongest uniqueness signal we can give each page at data level.

⚠ Alerts (this module has the sharpest rules): (a) "what locals choose" requires our real job history — for towns where we have few or no jobs yet (e.g., Tolland today), that row must be omitted or stated as town-housing logic, never as "our customers here pick X"; (b) all numbers must come from the census pull we already validate — no rounded-from-memory stats; (c) drive times are routing estimates, not magic — labeled as such.

Module 6 — Asbestos / lead-era demo note ("9×9 tile or black mastic in a pre-1980 floor = test before demo") SPAM RISK: MED (not spam — the risk is a different kind)

Buys: legitimately valuable homeowner safety info + huge differentiator in towns with old stock (New Britain 69% pre-1960, Hartford legacy areas). Homeowners ask this in trade forums weekly.

Search value: near-zero local competition answers it — clean informational win feeding the sale.

⚠ Alerts (and these are why I weigh it medium): (a) never imply we do asbestos removal — we test-and-refer or the homeowner's licensed abatement crew handles it; wording stays "suspect material gets tested before demo" full stop; (b) no scary "your house will kill you" framing — that's spam-psychology, banned by taste rule; (c) no dollar figures for abatement (the doc floating $8–12/sq ft — that's generic internet data, not ours, not published); (d) it goes only on pages where the housing math justifies it (pre-1980 stock), not carpet-bombed onto Avon colonials where it adds nothing.

So — does the full 6-module page cross into "AI slop at scale"?

No, and here's the honest proof: the page totals ~2,400 words, of which roughly 70% is combination-unique content (town data, service data, town+service facts, reviews) and 30% is deliberate company-standard boilerplate (process, warranty, license) — the same ratio as any credible national contractor site. Thin sites fail the reverse: 90% same words, 10% swapped town name. We fail that test only if we get lazy with the per-combo data — which is exactly what the data-gate rule (no data for a module on that page = module doesn't render) prevents.

What I will never do, even to make these pages "stronger"

• No invented prices, square-foot costs, or "average job" figures
• No credentials we can't verify (Schluter-certified etc.)
• No star-rating/review-count text until you give the real number
• No availability promises ("this week!") we can't guarantee
• No mass template waves without a demand reason for every page
• No sections that exist only to hold keywords — every module must answer a real buyer question or it goes

My recommendation

Build modules 1, 2, 4 in the pilot now (near-zero risk, highest payoff). Build 5 as the town-facts engine across all money pages (it's already half-built — census + permits live). Build 3 and 6 only where their alerts check out (symptom checker on repair/waterproofing/shower pages first; asbestos note on genuinely-old-stock towns: New Britain, Bristol, Hartford-legacy, Wethersfield — and only with the careful wording above).

Say "RUN deep-deck" and it gets built on bathroom-tile-remodeling-new-britain-ct in the preview with before/after shots again. Say "drop 3 and 6" if you'd rather keep the conservative set only.