ChatGPT Prompt to Rewrite a Pricing Page Without Guessing

1.9 percent is what Rivet Field’s pricing page has converted at for three quarters running, across roughly 38,000 visits a month. That is 722 trials, and more tellingly, 41 percent of sales calls open with the same question: “which plan do I actually need?” Of the trials that start on the wrong plan, 34 percent are gone within 60 days.

The page is not badly written. It is written from the inside, with feature names that mean something to the product team and nothing to a dispatcher with nine technicians and a broken spreadsheet. The prompt below rebuilds the page around the jobs buyers arrive with, then forces the model to flag every claim it cannot verify instead of inventing proof.

Reading the 1.9 Percent Correctly

Rivet Field sells scheduling and dispatch software to field-service companies, from two-van operations to depots with 30 technicians. Three plans have been on the page since 2023, and the names have never been tested.

  • Yard, 49 dollars a month. Two users, scheduling and invoicing. Chosen by 58 percent of trials and by almost nobody who stays past 90 days.
  • Crew, 129 dollars a month. Unlimited users, route optimisation, customer notifications. The plan the company wants to sell, and the one 22 percent of trials pick.
  • Depot, 349 dollars a month plus 12 dollars per technician. Multi-location, parts inventory, API. Requires a call, which 78 percent of qualified visitors never make.
  • The visitor mix. 61 percent arrive from a search for “field service scheduling software”, 22 percent from a comparison article, 17 percent direct or from a trial invite.
  • The objection log. 214 sales calls transcribed over six months, with 19 recurring objections. The top three are per-technician billing surprises, what happens at 12 technicians, and whether the annual discount is automatic.

Notice what is missing: no headline test, no button colour, no social proof strip. The page has a naming and sequencing problem, and testing button copy on a page that answers the wrong question just produces a cleaner wrong answer.

The Rewrite Prompt

This prompt asks for a page specification, not prose. You want the structure and the reasoning first, then the sentences, because a model asked for “compelling pricing copy” will produce three paragraphs of adjectives and a table that repeats your feature list.

You are a B2B SaaS pricing page strategist. Rewrite a pricing page for a
field-service scheduling tool. You are not allowed to invent proof, customers,
numbers, or awards. Where you need a fact you do not have, write [NEED: what].

Business context:
- Product: Rivet Field, scheduling, dispatch, invoicing for field-service teams.
- Customers: 1,900 paying accounts, from 2-technician shops to depots with 30.
- Current page converts 1.9 percent of 38,000 monthly visits into trials.
- 41 percent of sales calls begin with "which plan do I need".
- 34 percent of trials that start on the entry plan churn within 60 days.

Plans as they exist today:
- Yard, 49 dollars per month, 2 users, scheduling and invoicing.
- Crew, 129 dollars per month, unlimited users, route optimisation, customer
  notifications.
- Depot, 349 dollars per month plus 12 dollars per technician, multi-location,
  parts inventory, API, requires a sales call.

Buyer jobs observed in sales calls:
- "I need to stop double-booking two technicians for the same slot."
- "I need the office to see where everyone is without calling them."
- "I need to bill the same day, not on Friday from a shoebox of paperwork."
- "I need three branches on one schedule and one invoice run."

Recurring objections from 214 transcribed calls:
1. Per-technician pricing is opaque above 10 technicians.
2. Unclear what happens when a team grows past a plan limit.
3. Annual discount is not shown as a number.
4. No answer on data migration from a spreadsheet.
5. No answer on whether technicians need a paid seat.
6. No trial length or card requirement stated.

Produce, in this order:
1. Plan naming. Three names of one or two words that describe the size of the
   operation, not the feature set. For each, one sentence naming the buyer it is
   for and the job it solves. Do not reuse Yard, Crew, or Depot.
2. Page skeleton. Sections in the order a first-time visitor needs them, with a
   one-line purpose for each section.
3. Plan card copy. For each plan: a job statement under 12 words, exactly four
   feature lines in buyer language (not product language), the price display
   including the per-technician case, and one honest limit statement.
4. Comparison table. No more than nine rows. Row labels must be phrased as
   outcomes, not features. Mark any row where the honest answer is "not
   included" instead of shading it.
5. Objection block. Five objections answered in under 40 words each, using only
   the facts above. If an answer requires a fact I did not supply, write [NEED].
6. CTA set. One primary and one secondary label per plan. The entry plan must
   not require a call. Say plainly what happens after the click and whether a
   card is required.
7. Claims audit. A list of every sentence on the page that would need evidence
   before publishing, with the type of evidence required.

Rules:
- No "trusted by", no customer counts, no awards, no "limited time".
- No feature name longer than four words.
- Do not use "unlimited" unless a limit is stated in the same sentence.
- Do not describe the product as powerful, seamless, or all-in-one.
- If the three plans do not map cleanly to the four buyer jobs, say so and
  propose a different split before writing any copy.

The claims audit is the part to keep. It converts the model from a copywriter into a reviewer, and it gives you a list of blanks to fill from your own data rather than a page full of confident sentences nobody can defend in a sales call.

Six Inputs the Prompt Cannot Invent

The prompt above is only as good as the input block. Six fields decide whether the output is a rewrite or a generic template.

  • Actual plan limits. Users, technicians, locations, invoices per month. If Crew is unlimited users but capped at 12 technicians, that cap is the single most important number on the page and it is usually missing.
  • Billing mechanics. Per-technician charges, proration, annual discount as a percentage and a dollar figure. The page should show the arithmetic for an 11-technician shop, not describe the model.
  • Objection log. Real sentences from calls, at least 30. Nineteen recurring objections from 214 calls is enough; twelve objections from a brainstorm is not.
  • Buyer jobs. Four is a workable number. More than six and the plans stop mapping to them, which the prompt will tell you rather than hide.
  • Trial terms. Length, card requirement, what happens at the end. “Start free trial” without these is the most common honesty gap on a pricing page.
  • Migration reality. Whether you import from a spreadsheet, from a competitor, or not at all. If migration is manual, say manual and price the time.

One input you should deliberately withhold: revenue targets. A page written to hit a number this quarter will oversell the middle plan, and the churn shows up two quarters later.

What the Rewritten Page Looks Like

A good return for Rivet Field renames the plans around operation size, moves the per-technician arithmetic above the fold on the third card, and puts the growth question in the first objection block. The draft I would expect to see:

  • Plan names. Solo, Crew, and Multi-Site, each with a one-line buyer statement. “For two vans and one dispatcher” does more work than “essential features for small teams”.
  • Four feature lines per plan, in outcome language. “See who is free before you promise a slot” instead of “real-time availability engine”. The limit statement sits directly under the price: “Up to 12 technicians, then 12 dollars each”.
  • A nine-row table with visible gaps. Multi-Site genuinely lacks nothing important; Solo genuinely lacks route optimisation. Showing the gap sells the upgrade better than hiding it, because the buyer discovers it later anyway.
  • Objection block placed above the FAQ. The per-technician question answered with a worked example is worth more than six feature bullets, and it belongs before the comparison table, not after it.
  • Honest CTAs. “Start a 14-day trial, no card” on Solo and Crew, “Book a 20-minute setup call” on Multi-Site, with the call framed as setup rather than qualification.
  • A claims list with eight to twelve items. Anything from “cuts admin by half” to “most teams are live in a day” lands here, and most of it will not survive the audit, which is the point.

Read the result against your last twenty sales calls. If the page answers the first question in each call before the visitor books, the rewrite worked. If it answers questions nobody asked, the input block was thin, not the prompt.

Three Tests to Run Before It Ships

Do not A/B test the whole page at once. Three smaller tests isolate what actually moved. The objection miner below is the one to run first, because it produces the raw material for all three.

Here are transcripts or notes from [N] sales calls: [PASTE].

Extract every question a buyer asked before signing, in their own words.
Cluster them into at most seven themes. For each theme report:
- the exact phrasing used most often
- how many separate calls raised it
- whether it is a pricing, trust, migration, or fit question
- the one sentence that would answer it on a pricing page

Then rank the themes by how often they appeared before the buyer asked about
price, and flag any theme that never appears on our current pricing page.
  • Plan-name test. Keep everything else fixed and swap only the three names for job-based ones. At 38,000 monthly visits you have enough traffic for a two-week test at 95 percent confidence on a 1.9 percent baseline, provided you measure trial starts rather than page clicks.
  • Arithmetic-above-the-table test. Move the 11-technician worked example from the FAQ into the third plan card. This one usually lifts qualified trials rather than total trials, so measure 60-day retention alongside signups.
  • CTA honesty test. Change “Contact sales” to “Book a 20-minute setup call” on the top plan and see whether call volume rises while unqualified calls fall. If both rise, the qualification is happening on the call, which is a different problem.

A pricing page rewrite is mostly a decision about what to admit. The structure that forces admissions, stated limits, worked arithmetic, and an explicit list of unverifiable claims, is the structure that survives contact with a buyer who has already read two competitors. For the page-level framing around the table, the section order in our landing page prompt framework applies directly, and the test sizing rules in our A/B test copy workflow will stop you from calling a 0.2 point move a win on a 1.9 percent baseline.

Sources & Notes

Rivet Field, its three plans, the 1,900 accounts, and all conversion figures are illustrative. The prompt’s structure draws on standard pricing-page practice (job-based tier naming, visible limits, objection handling before comparison tables) rather than any proprietary method. Reviewed on 2026-11-02. Statistical claims about test duration are approximate and assume evenly split traffic; run your own power calculation before you trust a result, and re-check every price and limit against the billing system on the day the page publishes.

声明:1.本站所有文章,如无特殊说明或标注,均为本站原创发布。本网站中【文本生成/文生图】使用了DeeSseek人工智能生成内容技术, 算法来源说明:本网站未自行开发、训练或部署深度合成算法,所使用的人工智能算法来自:DeepSeek、Chatgpt、Google Gemini。 2.为了网站加载速度优化,本站预览图片进行压缩处理,所以有些不是太清晰,并不影响AI提示词生图、AI设计等内容。3.任何个人或组织,在未征得本站同意时,禁止复制、盗用、采集、发布本站内容到任何网站、书籍等各类媒体平台。如若本站内容侵犯了原著者的合法权益,可联系我们进行处理。--GUUNN.COM