Test run on www.texasfirstrentals.com: plan and timeline
Prepared by AnswerShare for Shane Wall and Stephen Marston, HOLT Group. Draft of 21 October 2026. Weeks are counted from the kickoff call. Calendar dates are set at kickoff.
What the test run is
One rewrite rule in your Next.js configuration (next.config.js) sends page requests from named AI crawlers, such as GPTBot, ClaudeBot and PerplexityBot, to AnswerShare. Vercel fetches the AI version of that page from us and returns it at the same URL. People, Googlebot, Bingbot and the other search crawlers never match the rule and get your site exactly as today.
What does not change: your DNS (www keeps pointing at Vercel), Vercel's firewall and bot protection, your site code, your language routing for /es-US, and your analytics. No secret or credential goes into your project. The rule itself, and why it sits in next.config.js, is on the rule page.
Timeline
| Phase | When | What happens | Who | Done when |
|---|---|---|---|---|
| 1. Kickoff with Stephen Marston | Week 0 Question for HOLT: Which date works for the kickoff call with Stephen? |
A technical call. We walk through the rule, the side-by-side pages, the preview tests and the data processing agreement, agree the test window and contacts, and make the access requests listed below. | Stephen Marston and HOLT engineers; AnswerShare engineering; Kevin Myers | Payment Received. Test window, contacts and access requests agreed. |
| 2. Your review of the rule | Weeks 0 to 1 | Your engineers review the rule and the crawler list, and remove any crawler you do not want served. Your counsel reviews the data processing agreement. In parallel we finish the AI versions of your English and Spanish pages and confirm the address the rule points to. | HOLT reviews; AnswerShare answers questions and makes requested changes | Rule text approved by HOLT, with a version number. DPA comments returned. |
| 3. Vercel preview and pre-launch tests | Weeks 1 to 2 | Your team opens a branch with the rule, which gives a Vercel preview. We run the pre-launch test matrix against it and share every result. Your team runs its own checks, including a form submission and the revert test. We also hold the webhook test session and take the "before" measurements (see measuring results). | HOLT creates the preview; AnswerShare runs the matrix; both review results | Every test in the matrix passes, and the before measurements are saved. |
| 4. Production enable on www | Week 2 Question for HOLT: Which weekday and time suits Stephen for the production deploy? |
Your team merges and deploys the rule to production in a deployment of its own, during a slot when Stephen and our engineer are both online. We repeat the core tests against www within the first hour. We confirm the slot 48 hours ahead. | HOLT deploys; AnswerShare tests | Core tests pass on www. AI crawlers receive the AI version; people and search crawlers receive the normal page. |
| 5. 7-day soak | Weeks 2 to 3 | We check every minute that search crawlers and browsers still get your normal pages, that AI crawlers get the AI version, and that error rates and response times stay inside the limits below. Our AI-crawler traffic watch alerts us if AI-crawler traffic stops. | AnswerShare monitors; HOLT can ask for a revert at any time | Seven consecutive days with no stop condition. |
| 6. Handover call | The first day after the 7-day soak ends; the date is fixed once the production slot is agreed | We walk your team through the AnswerShare dashboard and give dashboard access on this call. Access is given only at the handover, after the soak. | Both | Your named users can sign in to the dashboard. |
| 7. Baseline-vs-result readouts | At the handover call; 30 days after enable; 90 days after enable [CONFIRM: cadence] | Each readout compares results with the baseline taken before launch, using the same instruments. The first readout covers the crawler side: AI crawler visits by type, the user-intent share, and the size and cost of what AI crawlers receive. The 30-day and 90-day readouts add answer share, the GEO and SEN scores, and AI-referred sessions from Google Analytics. | AnswerShare prepares; both review | A written readout with every figure, its date and its source. |
Access we ask for at kickoff
We ask at kickoff, before the preview, so that the "before" picture is complete. All access is read-only and goes to aria@aryah.ai.
| Request | Why | Notes |
|---|---|---|
| Google Analytics 4: Viewer role on the Texas First Rentals property (property level only, not Editor, Admin or account level) | AI-referred sessions by assistant, before and after launch. Google Analytics keeps history, so the before period can be read once access exists. | We found Google Tag Manager on your site but could not see the GA4 property ID, so we need you to name the property. We read aggregate reports only: no revenue, transactions, conversions, funnels or user-level data. |
| Google Search Console: Restricted user on www.texasfirstrentals.com | Search queries and impressions on the Google side, which the rule does not touch. They show what changed for reasons other than the rule. | We could not see a Search Console verification record. If the property is verified another way (for example through Google Analytics or Tag Manager), please add us as a Restricted user. If it is not set up, tell us and we will plan without it. |
| 30 days of Vercel request counts by AI-crawler user agent | The only "before" count of AI crawler visits. We have no view of your traffic until the rule is live. | Counts only, no IP addresses. How far back Vercel keeps request logs depends on your plan and settings. If no history exists, we report crawler visits from launch onward and say so. Question for HOLT: How many days of request logs does your Vercel plan keep, and can your team export counts by AI-crawler user agent for the last 30 days? |
| Preview access for testing | Vercel previews are often protected by Vercel Authentication. | Either your team runs our test script, or you give us a protection-bypass value on a call (never by email) for the test window. Question for HOLT: Which do you prefer: your team runs our test script, or you give us a protection-bypass value on a call? |
| Keyword research exports, if you have them | To build the answer-share prompt set from your own data rather than guesses. | Two views help: pages with their keywords and search volume, and keywords mapped to pages. Any tool's export works. |
Content updates
Content updates reach us by webhook. When Vercel deploys a new production version of your site, a short step in your deploy sends us the addresses of the pages that changed, signed with a secret we give you on a call. We then refresh the AI version of those pages from your public site within minutes. We do not poll or crawl your site on a schedule to find changes. Our step-by-step publish webhook guide (version 1.0, 8 October) covers the set-up, and we test it together during phase 3. Stephen receives the guide before the kickoff call.
Our page fetcher identifies itself with its own user agent, which the rule does not match, so a refresh always reads your normal page. We give you that user agent and its IP address before the webhook test so you can recognise it, or allow it in your Vercel firewall. [CONFIRM: fetcher user agent and IP]
What we need from HOLT
Each item links to the place on these pages where we ask it. Answers on a call or by reply are fine.
- Kickoff date (plan and timeline, phase 1)
- Production deploy weekday and time (plan and timeline, phase 4)
- Vercel log retention and AI-crawler counts for the last 30 days (access we ask for)
- Preview access: your team runs our test script, or a protection-bypass value on a call (access we ask for)
- Stop limits for AI-version failures (stop conditions)
- Stop limit for page speed (stop conditions)
- Final list of pages under test (preview testing)
- Firewall and bot protection settings on the preview (preview testing)
- Any account or transaction paths on www (the rule)
- Your Next.js version (the rule)
- Whether you use Instant Rollback on your Vercel plan (the rule)
FAQ review comes after go-live
The FAQ review is a post-launch task. Nothing in phases 1 to 4 waits on it. Until you approve the FAQ, the AI version carries only FAQ content already published on your site, never the draft. The FAQ document we sent on 18 October moves to this step.
Stop conditions and revert
The rule is revertible in one step: roll back to the production deployment before the rule (Vercel's Instant Rollback, no rebuild), or remove the rule and redeploy. Because the rule ships in a deployment of its own, a rollback removes nothing else. Details are on the rule page.
| Condition | Action |
|---|---|
| Any browser or search crawler receives the AI version of any page | Revert at once, then investigate. |
| The AI version fails (5xx or timeout) for more than 1% of AI-crawler requests in any hour, or for 15 minutes in a row Question for HOLT: Do these limits (more than 1% failures in an hour, or 15 minutes in a row) suit Stephen's team, or would you set different ones? | Revert, or fix within the hour if the cause is known and on our side. |
| Your normal pages get slower for people after the rule ships (median time to first byte more than 10% above the week before) Question for HOLT: Is a 10% rise in median time to first byte the right trigger for your team, or would you set a different limit? | Revert and investigate. |
| HOLT asks for a revert for any reason | Revert. No justification needed. |
Who does what
| HOLT / Texas First Rentals | AnswerShare |
|---|---|
| Reviews and owns the rule; removes any crawler name it does not want. | Writes the rule and the crawler list; sends any change as a pull request. |
| Creates the Vercel preview; runs its own checks, the form test and the revert test. | Runs the pre-launch test matrix and shares every raw result. |
| Deploys to production in the agreed slot; can revert at any time. | Builds and hosts the AI versions of your pages; keeps them current from your webhook. |
| Adds the webhook step to your deploy. | Monitors every minute and alerts on a stop. |
| Grants read-only analytics access. | Takes the baseline and writes each readout. |
| Reviews the FAQ after go-live. | Signs the data processing agreement; never receives data about your visitors. |