The verdict · council-ratified, 15 September 2026

Our own August edits cost more than any competitor did. We freeze what earns, reverse three changes once, fix one setting, and spend the 90 days making the money pages answer-shaped.

No slug, title or redirect consolidation happens before January. Everything waiting on the client moves to a dated blocked column, off the scoreboard. Nothing below is a promise of recovery; each move has a measured exit gate.

Survey says · the five moves of the next 90 daysreviewed by an 11-seat council

Freeze the top 50 earners in the tracker tab the content system reads. Umer-initials column. The 15 scheduled articles pause until each clears a money-page collision check.

today · 2 h

Reverse the three August changes once — teacher-appreciation slug, awards title, guessing-games slug — then never touch those URLs again.

day 1 · 1 h

Turn off the one setting shipping 1 MB pages (WP Rocket Remove Unused CSS), purge, re-measure. Money pages under 300 KB.

day 2 · 0.5 h

Make the money pages eligible — dedupe the NYC article, index the four new pages plus NJ/NYC, cut the contact form to five fields with the price and phone above it.

week 1 · 5 h

Make them answer-shaped for Q4 — answer blocks and FAQ on seven money pages, one truthful cost guide, the Halloween post refreshed and linked before the season closes.

weeks 2–4

Clicks · last 3 mo

1,360

was 1,770 · −23% · GSC 15 Sep

Commercial clicks / 90d

from ~103K impressions · 0.05% CTR

Avg position

43.5

was 40.8 and worsening

Mobile CWV good

0/143

pages ship 0.6–1 MB of HTML

Domain Rating

−25 in 6 months · Ahrefs 15 Sep

AI crawlers

403

GPTBot · ClaudeBot · PerplexityBot

Measured observed directly on 15 Sep — source named at the claim Modelled an estimate, labelled so — never a promise

Why this must happen now

Q4 is the buying season, and the site enters it wounded by its own programme.

Corporate holiday bookings are decided in October and November. Halloween queries already sit at positions 14–22 with 1,600 impressions. Every week of delay hands those enquiries to GigSalad vendors.

The deeper problem is control, not effort. Since June the programme shipped a 335-post rewrite, 376 re-dates, four new money pages and 18 commercial articles. In the same period it also deleted two of the site's own protected earners and cut the top post's clicks by two thirds. Output is not the constraint. Protecting what works is.

Proof · the diagnosis is measured, not argued

Three facts decide the whole plan.

1 · The programme deleted its own earners Measured

URL · what changed	Date	Before	After	Source
games-for-teachers-appreciation-day → re-slugged to a different topic	28 Aug	135 clk/qtr · pos 11.8	0 clk · gone	GSC compare, 15 Sep
funny-employee-appreciation-awards → title moved off "101 funny employee awards"	20 Aug	584 clk/qtr	208 clk/qtr	GSC compare, 15 Sep
guessing-games → guessing-games-for-work	28 Aug	pos 9.9	pos 41	GSC compare, 15 Sep

Modelled Reversing all three returns about 511 clicks a quarter against a year-on-year hole of roughly 8,600. That is 5%: trust repair, not recovery. A 30 October checkpoint decides whether those URLs are then left alone for good. The wider collapse predates the June programme and is not promised back by any date.

2 · Page weight is one setting Measured

curl · rendered HTML bytes · 15 Sep 2026

/game-show-team-building/ 964 KB (wpr-usedcss inline block: 802 KB)
/team-building-activities/nyc/ 1,049 KB (wpr-usedcss: 811 KB)
/ (homepage) 599 KB (wpr-usedcss: 396 KB) · cf-cache-status: DYNAMIC on all three

WP Rocket's Remove Unused CSS inlines the entire "used" stylesheet into every page. Switching it off is a single setting with a one-click rollback. Expected result is money pages under 300 KB; the regression check runs at 390, 768 and 1366 px before it is called done.

3 · The demand exists and page type, not authority, is the blocker Measured

GSC 90d · Ahrefs 15 Sep

Commercial + local queries: 103K impressions → 52 clicks · NYC page 32,263 impr / 0 clicks / pos 68 · NJ page 11,420 / 3 / pos 63
"team building events nj" 1,600/mo · KD 0 · site absent | content gap vs 3 real vendors: 3,088 keywords
July diagnosis: a DR-4 competitor outranks this DR-18 site on the money terms — page type decides, not authority

The site already wins where the page type fits: 2 of the top 5 for "game show company NJ", 3 of 5 for "team building game shows NYC" and "school assembly game show", and the AI Overview cites the site's own cost maths. The plan extends what is proven to work instead of guessing.

How this was verified

Evidence trail · all 15 Sep 2026

crawl: 483/483 sitemap URLs fetched · 12 broken internal links re-fetched individually · 2 candidate findings DISCARDED on inspection (a schema heuristic, a Cloudflare decode URL) — reported as non-findings, not defects
GSC: 3-month + compare exports · coverage drilldown · URL inspection on the money pages
AI access: curl -A per agent → GPTBot 403 · ClaudeBot 403 · PerplexityBot 403
the 69-starts/8-submits form figure could NOT be re-verified (no GA4 access) — treated as unverified and replaced by a WPForms-entries baseline set in week 1

The control case · proving we stay quiet

The plan's first KPI is the one where nothing happens.

A skeptical reader should test the opposite claim: will this programme stop breaking what earns? The freeze makes that checkable.

The protected set

Top 50 URLs by 90-day clicks

Exported into the tracker tab the content system reads and marked PROTECTED. Baseline: the 13 core earners carry 1,082 clicks/90d.

The mechanism, not a rule

An initials gate, inside the workflow

No title, slug or re-date change on a protected URL without a before/after row and Umer's initials. August happened because the rule lived in a document; this one lives where publishing happens.

The queue check

15 scheduled articles paused

Each returns to the queue only after a title/slug collision check against the 7 money pages. The 5 Sep NYC article that duplicated a money page's exact title is why.

Read it weekly in one number: unapproved changes on the protected list. Target every week of the 90 days: 0. The first entries in the log are the three reversals themselves, initialled, with their 30 Oct checkpoint attached.

The plan · gates, not calendars

Three gates. Nothing advances until its gate passes, and every gate is checkable by anyone.

Week numbers are working estimates; the gate criteria are the commitments. About 12 hours of work sits before gate 1, inside a 25-hour weekly cap.

Gate 1 · The bleeding has stoppedtarget: end of week 1

Freeze mechanism live in the tracker; 15 scheduled articles set to Draft pending collision checks

2 h

Three August reversals applied once, logged with initials

1 h

WP Rocket Remove Unused CSS off, purge, PageSpeed re-test; Cloudflare untouched in the same change

0.5 h + QA

Client decision email sent by Umer with a 1 October deadline

1 h

Exit gate — all four, measured Tracker tab has 50 protected rows and initials on the rule row · publishing queue shows 0 scheduled posts · old slugs return 200 and new slugs 301 back · one money page measured under 300 KB with no layout regression at 390/768/1366.

Gate 2 · The money pages are eligibletarget: day 14

NYC dedupe: only one URL carries the NYC title and H1; indexing requested on the four money pages plus NJ/NYC

1 h

Contact form cut to name, email, phone, event date, group size — "Shows from $2,495 · call (201) 357-2979" directly above it

3 h

Test submission to umer.tsp@gmail.com only. WPForms entries Jun–Sep become the enquiry baseline, replacing the unverified analytics figure.

Schema hygiene: 3 JSON-LD parse errors fixed, duplicate Organization removed; nav "New York" repointed; 12 redirected URLs out of the sitemap

3 h

Exit gate 4 of 4 money pages indexed by day 14 (GSC URL inspection) · form live on desktop and mobile with the baseline number in the KPI sheet · 0 unparsable structured-data items · 0 redirecting sitemap URLs.

Gate 3 · Answer-shaped for Q4target: day 30 · hard stop for Halloween: 31 Oct

Answer blocks + FAQ on the seven money pages — content only, no URL, title or H1 changes

2 days

Each page states what, who, group size, duration, price and area in its first 80 words. Price and group-size lines await Lisa's confirmation.

One truthful cost guide behind the AI Overview citation, with the $2,495–$7,875 ladder; both existing cost posts aligned to it

1 day

Halloween post refreshed for 2026 with a hosted-show block; contextual links from the 10 largest holiday posts

5 h

NJ page answer block for "team building events NJ" — URL, title, H1 unchanged; four striking-distance retitles, each with a before/after log row

1 day

Exit gate Numbers identical across /pricing/, the guide and both posts · each money page answers the buyer's five questions in its first 80 words · Halloween queries re-inspected before 31 Oct · every retitle has its log row.

Then a standing Friday rhythm: the blocked-column report (item, requested date, days open, owner) and the Q4 2027 KPI column — because the consolidation parked to January pays off next holiday season, not this one.

Strategist concurrence · AEO Agent content pipeline, 15 Sep

The system that writes this client's content was shown the plan. It agreed, and it surfaced three collisions already sitting in its own queue.

The plan was reviewed with the AEO Agent strategist that runs the It's Play Tyme content pipeline, in the client's own thread. It endorsed the three reversals and the January parking without contradiction, and it produced the two things execution needs.

It confirmed	Detail
3 HIGH-collision articles are already queued	One is already fully written: a "Corporate Holiday Party Entertainment Companies in NJ" piece that shares the holiday money page's head. The large-groups NJ/NYC article hits both geo money pages at once — the exact 5 Sep failure, about to repeat. This is the freeze earning its place before it is even switched on.
The freeze binds in the pipeline, not the tracker	The system obeys the client's overrides file at generation time and never reads the tracker tab. The protected list and the collision rule must live in that file to bind it. The tracker stays as the human audit trail. It drafted the exact block to add.
A silent write bug on the calendar	Field edits can return success and write nothing. Every change is confirmed by reading the row back, never by the success response. This is now the logging rule for the whole freeze.
WPForms alone will undercount	Its estimate puts most enquiries on phone, GigSalad and referral, not the website form. WPForms stays as the always-readable floor number; the channel tally and call tracking ride the client email. Labelled an estimate on its side too.

The one method correction we adopted: restore the exact original slugs Google already indexed, never coin a third corrective slug. A revert to a known URL re-consolidates fast; a new slug is a second migration and adds the very volatility the reversal undoes. Two objections were accepted with modification (WPForms as a floor not the funnel; the author page moved from cut to parked), and one declined: new comparison URLs wait for January, because shipping net-new URLs during the freeze window is the exact behaviour the freeze exists to stop.

Full exchange archived with the strategy files. Nothing in the pipeline has been changed; the strategist is holding a row-by-row Draft-and-purge list for sign-off.

Independent strategist read · research-first, 16 Sep

We then made the content system re-examine the site on its own, fed it none of our conclusions, and cross-questioned what it found. It confirmed our core, added four findings, and disputed two of our moves.

The first review handed the strategist our plan. This one did the opposite: it ran its own on-box data reads and live page inspection, and we only challenged its answers. An untainted second opinion is worth more than an echo.

It confirmed, from its own data

Finding	What it measured independently
Page weight	Fetched the money pages itself: 964 KB to 1.05 MB. Matches our reading.
Buried commercial demand	NYC page 31,600 impressions at position 68, NJ page 11,130 at position 74, near-zero clicks. The biggest impression sources on the site, invisible on the SERP.
Estate shape	88.6% blog, a thin fully-indexed page layer, nothing dropped out of the index. The problem is ranking erosion, not de-indexation.
AI crawlers blocked	Confirmed 403 for the AI bots. Then refined it, see below.

Four findings we did not have

New · pricing Measured

Three different prices are live at once.

The pricing page shows a $2,975 floor, the service pages show a $2,495 to $7,875 ladder, and the holiday page shows $2,975 to $5,475. Separate from the known prizes contradiction. Its read: the $2,975 floor sits on the newest pages, so the older ladder is likely stale. Settle it from the live quote flow, the page dates, or Lisa.

New · technical Measured

The AI block is a firewall rule, not robots.txt.

Googlebot gets 200, the AI bots get 403, and robots.txt itself serves 200 to GPTBot. So the fix is a Cloudflare firewall change, not a robots edit. This changes the exact ask to the client.

New · structure Measured

The corporate head is the worst cluster.

18-plus competing live URLs, including a triple duplicate of the corporate page plus children and a holiday duplicate pair. Worse and more specific than our game-show-cluster framing.

New · geo + brand Measured

Pages we wrote off are live and pulling traffic.

Philadelphia, Connecticut and New York state geo pages carry 700 to 1,700 impressions each, which contradicts our "Pennsylvania removed, Connecticut non-viable" note. Reconcile before any geo consolidation. It also caught "PlayTyme" rendered with a capital T against the brand rule.

Where it disputes our plan Its recommendation, weigh it. Two of its explicit do-NOTs push back on our moves: it says do not rebuild the money pages for thin content, because they measured deep and schema-rich, and do not build more NJ city pages, because that cluster is a doorway pile at its ceiling. On Q4 it argues the gap is routing and authority, not content, so the win this quarter is to link the ranking seasonal blogs into the buried money pages now, rather than build.

Its priority order for the 90 days

Do, in order	Do not
1 · Unblock AI crawlers at Cloudflare (the firewall rule)	Rebuild the money or geo pages, they are already deep and schema-rich
2 · Internal-link the buried geo and holiday pages from the high-traffic ranking blogs	Build more NJ city pages, the cluster is a doorway pile at its ceiling
3 · Consolidate the corporate cluster to one canonical per head	Add more team-building-head blogs, 18-plus already compete
4 · Reconcile pricing to one source so the schema is trustworthy	Prioritise the page-weight work ahead of the crawler unblock and consolidation, Googlebot still gets 200
5 · Fix the brand-name case while touching these pages, AI quotes titles verbatim

Full independent read archived with the strategy files. The strategist changed nothing on the site or in the pipeline; these are findings and recommendations for our decision.

Scope · no blanks

Everything from the 32-task backlog has exactly one home.

Disposition	What	Why
In the 90 days	Freeze · 3 reversals · WP Rocket · indexing + dedupe · form fix + WPForms baseline · client email (sender: Umer) · Halloween + holiday links · answer blocks + FAQ · cost guide · retitles + schema + nav/sitemap cleanup · citation audit	Each has a first step we can take without the client and carries no slug, title or 301 churn on ranking URLs.
Parked to January	All consolidation: the 22-URL game-show cluster, the NJ/Hackensack merge, the 452-row prune, NYC listicle, data study, format-page rebuilds, category consolidation	Consolidation is the August failure mode, and nothing in it ranks before November. Prerequisites to reopen: 4/4 money pages indexed · initials gate live 30 days · Lisa's rulings received · enquiry channel split known.
Blocked, dated, off the scoreboard	Cloudflare AI-bot unblock (requested 21 Aug — oldest clock, 25 days) · GA4 access · GBP manager access · edge caching · 404 catch-all sign-off	These are requests, not tasks. Dating them shows whose clock is running; the Friday report reads the ages out.

Cut from the 90 days — struck, with the reason on the card

Reddit programme + AI prompt panel. 124 comments drafted, 0 postable until accounts and the crawler unblock exist. Hours returned to the money pages.

Author entity page. Needs the client's photo and bio; no first step this month.

Venue partners and paid memberships. Real budget for a lever that pays after the season. Deferred, money saved this quarter.

"CWV 0 → 80" as a KPI. A developer deliverable with no developer. Page bytes and lab LCP on seven money pages replace it.

Findings worth money · including the ones that cost less

Four findings lower the bill before anything raises it.

Saves cash Measured

Do not pay for spam-link "removal".

The 12–15 Sep burst of SPAM-tagged domains looks alarming and is worth $0 to fix. Google discounts these; a disavow is filed only past 100 domains or on a manual action. Weekly monitor, no vendor.

Saves spend Measured

Connecticut is not a market.

Every planned CT city page measures under 10 searches a month. The pages the old plan would have built are not built.

Free recovery Measured

The biggest wins are reversals, not builds.

Three restores from WordPress revisions and one cache setting. Zero new content spend sits in front of gate 1.

Honest metric Replaces an unverified one

Enquiries counted where they actually land.

The quoted 69-starts/8-submits figure could not be verified. WPForms entries are counted directly in wp-admin, free, from week 1.

The one decision we need

Four yes/no answers by 1 October. Each has a stated if-no branch.

Client email · sent by Umer · reply by 1 Oct 2026

The asks, with what happens on silence

1 · Unblock AI crawlers at Cloudflare (ai-train=no may stay).

If no — the AEO track stays parked and its blocked-clock is read out every Friday; ChatGPT, Claude and Perplexity keep recommending competitors.

2 · Google Business Profile manager access.

If no — the homepage keeps ranking 1–4 for NJ terms with 0 clicks while the map pack owns the screen.

3 · GA4 access for key-event setup.

If no — WPForms entries remain the only enquiry metric; phone and email enquiries stay invisible.

4 · Lisa's rulings: prizes included? group-size floor 10–500 or 20–300? virtual offered? average enquiry reply time?

If no — answer blocks and the cost guide ship with only the facts already published on the site, and say so.

One fact we tally ourselves this week: the channel split of the 232 recorded enquiries. If under 30% came through the website, the email leads with GBP and call tracking instead of Cloudflare. Week 1 is identical either way.

Acceptance · how the 90 days are judged

Eight numbers, each readable today by one person with the sources named.

KPI	Baseline · 15 Sep	Day 90	Read from
Unapproved changes on the protected list	—	0	Weekly tracker diff vs the 50-row list
Money pages indexed	1 of 4	4 of 4 by day 14	GSC URL inspection
Commercial + local clicks / 90d (fixed query regex)	52	100	GSC query export
WPForms entries / month	set week 1	+50%	wp-admin → WPForms → Entries
The three reversed posts	0 / 208 / 0 clicks	50% of prior clicks at 30 Oct, else left alone	GSC page report
Money-page HTML · lab mobile LCP	964 KB–1 MB	<300 KB · <2.5 s	curl bytes · PageSpeed lab
Blocked-column age	oldest 25 days	reported every Friday	Tracker
Q4 2027 column · holiday + NJ/NYC clicks	opened now	tracked monthly	GSC page filter

Modelled A plausible day-90 report reads: protected-list changes 0 · indexed 4/4 · commercial clicks 52 → 91 · WPForms 3/mo → 5/mo · NYC page position 68 → 44. Plausible, labelled as such, and not a promise. A return to last year's total traffic is not the goal and is not promised.