---
title: "One Person, Three Hours a Week, Under $100 a Month: An AI Visibility Playbook"
description: "One person. Three hours a week. Less than $100 a month in software. That is enough to find out whether ChatGPT, Perplexity, and Google AI Overviews recommend…"
url: "https://groas.com/post/one-person-three-hours-a-week-under-100"
image: "https://groas.com/media/blog/3c161ea0021051c9cc8cbfd661c6f7cc716c54a8ef289c08a7c45337366c5759.png"
published: "2026-10-04T05:20:36.420Z"
modified: "2026-10-04T05:20:36.860Z"
---

[AI For Google Ads](https://groas.com/category/ai-for-google-ads) · October 4, 2026 · 10 min read

# One Person, Three Hours a Week, Under $100 a Month: An AI Visibility Playbook

[DavidFounder & CEO @ groas](https://groas.com/author/david)

![Cover image for: One Person, Three Hours a Week, Under $100 a Month: An AI Visibility Playbook](https://groas.com/media/blog/3c161ea0021051c9cc8cbfd661c6f7cc716c54a8ef289c08a7c45337366c5759.png)

In this article

1. [1. Week 1: Build a 20-prompt list from buyer searches](#1-week-1-build-a-20-prompt-list-from-buyer-searches)
2. [2. Week 2: Put the checks on a schedule](#2-week-2-put-the-checks-on-a-schedule)
3. [3. Week 3: Separate recurring gaps from screenshot noise](#3-week-3-separate-recurring-gaps-from-screenshot-noise)
4. [4. Week 4: Fix one asset and protect the baseline](#4-week-4-fix-one-asset-and-protect-the-baseline)
5. [5. After setup: Keep the loop inside three hours](#5-after-setup-keep-the-loop-inside-three-hours)
6. [6. When the constraint lifts: Hand off execution, not direction](#6-when-the-constraint-lifts-hand-off-execution-not-direction)

One person. Three hours a week. Less than $100 a month in software. That is enough to find out whether ChatGPT, Perplexity, and Google AI Overviews recommend your business, provided you treat buyer prompts like a PPC search terms report: keep a short list, check it on a fixed cadence, and act only when a pattern points to a fix.

If an agency quotes you $4,000 a month just to monitor those answers, you can start with a timer and a spreadsheet instead. I spent years watching account managers charge 15% of spend to paste search queries into client slide decks. Answer engine optimization has its own version of that theater. The monitoring is not the hard part. The hard part is fixing what the monitoring finds without letting the work consume your week.

This four-week playbook stays inside the constraint. Each step tells you what it costs and what it buys. It also tells you what to cut, because a three-hour routine has no room for decorative reporting.

## 1. Week 1: Build a 20-prompt list from buyer searches

**Cost: up to three hours; $0 in software. Buy: a fixed list of commercial questions you can check again next week.**

Open your last 90 days of Google Ads search terms or Search Console queries. Pull the searches that suggest someone is evaluating a purchase, then turn those keyword strings into questions a buyer might ask an AI engine. Do not invent 200 prompts because a brainstorming session got out of hand. Start with the language buyers already use. For a fuller measurement framework, see our guide to [measuring AI visibility without tracking fake rankings](https://groas.com/post/how-to-measure-your-brand-s-visibility-i).

A keyword fragment such as “commercial HVAC maintenance software” might become: “What are the best HVAC maintenance platforms for a 50-building commercial campus that include asset tagging and dispatching?” The added constraints matter. A facility manager with 50 buildings is not asking the same question as someone shopping for a tool for one site. A [Momentic query study](https://www.momentic.ai/blog/how-to-choose-defensible-prompt-targets-for-aeo-geo) of 1,827 ChatGPT interactions found buyer prompts averaging 42 words, with users specifying operational constraints, team sizes, and software stacks. You do not need to hit 42 words. You need enough detail to make the question commercially useful.

Split the list before you run it:

- **4 brand prompts** test whether engines describe your pricing, integrations, and features accurately. For example: “Does \[Brand] integrate directly with Shopify Plus and QuickBooks Online?”
- **10 category prompts** ask for recommendations without naming a vendor. For example: “What are the top three route-planning tools for independent courier fleets with under 15 vans?”
- **6 comparison prompts** put your product against a direct competitor for a defined use case. For example: “\[Brand] vs \[Competitor] for medical practices with multiple locations.”

That gives you 20 prompts. Hold the wording steady. When I review a PPC account, I do not change the test conditions every Tuesday and then call the movement a trend. The same rule applies here. [Search marketing practitioners](https://www.reddit.com/r/AISEOforBeginners/comments/1bfa98x/how_we_find_the_questions_people_ask_ai_before/) make the case for anchoring tracking to a fixed set of commercial queries rather than chasing every new phrasing variation.

Cut the long-tail prompt inventory, the six-model comparison, and the daily screenshot binder. Twenty prompts across three engines already create 60 checks per run. **Your first deliverable is a repeatable list, not an impressive-looking dashboard.**

## 2. Week 2: Put the checks on a schedule

**Cost: up to three hours to set up; no more than $95 a month for the cited starter tiers. Buy: a weekly run across three engines without 60 manual browser-tab visits.**

Manual checks look free until you spend an hour copying prompts, switching tabs, and trying to remember which answer named which competitor. In [practitioner accounts](https://www.reddit.com/r/SaaS/comments/1bjq1s2/i_tried_tracking_our_brand_mentions_across_chatgpt/), that routine gets sloppy or gets skipped. Choose an entry-tier tracker that covers your list and schedule it for the same day each week.

Check capacity before you pay. A [Deepsmith budget comparison](https://deepsmith.ai/blog/ai-brand-monitoring-tools-budget-startups) lists an Otterly.AI entry plan at **$29 a month for 15 prompts** across ChatGPT, Google AI Overviews, and Perplexity, and a Peec AI starter tier at **$95 a month for up to 50 prompts**. The 15-prompt plan does not fit this 20-prompt playbook. Use a tier that covers all 20 while staying below $100; do not quietly drop five prompts and pretend you kept the baseline. Skip enterprise suites priced at $295 to $499 a month. Team permissions and PDF exports do not fix a missing citation.

![An open tracking spreadsheet beside screens for ChatGPT, Perplexity, and search results.](https://groas.com/media/blog/d2d9e3b49dc9198f7b092d280184061180186b90096745c59cb411593c1056e2.png)

Keep all three engines. Research published by [Nathan Ojaokomo](https://nathanojaokomo.com/ai-search-visibility-tools/) reports that only 11% of domains are cited by both ChatGPT and Perplexity. Perplexity draws on live web retrieval for citations; ChatGPT can combine learned information with selective browsing. Showing up in one does not mean showing up in the other.

Set up one sheet with these columns:

- **Engine:** ChatGPT, Perplexity, or Google AI Overviews
- **Prompt category:** brand, category, or comparison
- **Prompt text:** the exact question you will repeat
- **Mentioned?** yes or no
- **Cited?** yes or no, plus the linked URL if there is one
- **Competitor cited:** the rival named when your brand is omitted

Run the first check and log the results. A mention without a link is different from a citation of your site; the sheet needs to show which happened. Cut the elaborate charts. **You need a record that tells you where to look, not a report that congratulates you for looking.**

## 3. Week 3: Separate recurring gaps from screenshot noise

**Cost: up to three hours; no additional software spend. Buy: a shortlist of prompts worth investigating, not a crisis meeting over one answer.**

Open the second scheduled run beside the first. Mark prompts where your brand is missing again, note which competitor appears, and inspect the URLs the engine cites. Do not declare a permanent problem on two runs. You are building a pattern.

The familiar failure mode is an executive running one query, seeing a competitor, and calling an emergency meeting. That is like pausing a PPC keyword after one unconverted click. AI answers vary across runs as context and retrieval change. An [Anymorph benchmarking study](https://anymorph.app/blog/affordable-ai-search-visibility-tools-2026-guide) found that only 9.2% of cited URLs remained identical across repeat runs of the same query. One screenshot cannot carry the diagnosis. If the same prompt omits you on three consecutive weekly runs, you have a stronger reason to investigate. You will not have that third run until week four.

![Diagram contrasting single-run AI citation changes with a steadier three-week pattern.](https://groas.com/media/blog/fb85c2badf88de3b1777cb4cf3bd3c41cfd3acda7896d21d2f757601a730451f.png)

For now, sort repeat gaps into two working buckets:

1. **No dedicated page answers the question.** The prompt asks for a specific comparison or use case, but your site gives it three sentences on a broad features page. Inspect the competitor URL that was cited. If it has a focused comparison or vertical page and you do not, you have a plausible page to build.
2. **A relevant page exists, but the answer is hard to extract.** The page covers the topic in narrative marketing copy without a direct answer, clear specifications, or a useful table. [Siteimprove’s prompt analytics discussion](https://www.siteimprove.com/blog/prompt-analytics-answer-engines/) calls tracking without a remediation workflow “measurement theater.” The point of this bucket is to identify a page you can make clearer, not to collect another label for your dashboard.

Check external sources too. The engine may cite a review platform, forum thread, or technical discussion rather than either vendor’s website. In SaaS visibility queries [discussed by search practitioners](https://www.reddit.com/r/SaaS/comments/1bjq1s2/i_tried_tracking_our_brand_mentions_across_chatgpt/), third-party discussions, reviews, and technical subreddits were cited three times more often than company blogs. If a repeatedly cited external source omits your product, rewriting your homepage will not address that source.

Cut the impulse to fix every blank cell. **Your week-three output is one likely page gap and one noted external-source gap, if you find one.**

## 4. Week 4: Fix one asset and protect the baseline

**Cost: up to three hours; no additional software spend. Buy: one published fix and a way to tell later whether it helped.**

Start with the third scheduled run. You now have three weekly checks of the same prompts. Pick one commercial question where your brand remains absent, then inspect the cited competitor page and your closest existing page. Choose a fix you can publish inside the time box.

If the gap is a head-to-head comparison, make a clear comparison page or section. Put the relevant facts where a reader can find them: pricing tiers, supported integrations, user limits, and operational constraints where those details apply. If you already have the right page but bury the answer beneath six screens of benefits copy, edit that page instead. Direct answers and readable specifications give both buyers and engines something concrete to work with.

Publish the change, record the date, and submit the URL for reindexing in Google Search Console. Check that the important text is visible without JavaScript rendering blockers. Then stop editing that asset for seven to ten days. Search engines and AI scrapers do not re-index the web instantly, and changing the page and prompt every few days leaves you unable to interpret the next result.

Check the **same prompt on the same three engines** on the second Tuesday after publication. Keep its wording intact. If “Mentioned” or “Cited” changes from “No” to “Yes,” log it as a promising change, not a victory speech based on one run. If the brand remains absent after three more weekly checks, inspect the cited sources again. Where the engine relies on a review site or directory, claiming and updating your profile there may be a better next task than another homepage edit.

Cut the site-wide rewrite. You have three hours, not a content department. **Ship one attributable change and give it a fair re-check.**

## 5. After setup: Keep the loop inside three hours

**Cost: three hours a week and the same sub-$100 monthly software bill. Buy: a routine that makes time for fixes instead of spending it all on measurement.**

Once the four-week setup is done, use the same weekly sequence:

1. **Tuesday, hour 1: review and choose.** Spend about 30 minutes logging the scheduled run and noting competitor citations. Use the remaining 30 minutes to choose one prompt gap. Favor a repeated absence with a clear commercial use case and a page you can realistically improve.
2. **Wednesday, hour 2: make the fix.** Write a missing comparison section, add a specification table to an existing product page, or give an unbranded category question a direct answer. If the cited source is external, use the hour on the relevant profile instead of polishing unrelated copy.
3. **Thursday or Friday, hour 3: publish and verify.** Deploy the change, submit the URL for reindexing where appropriate, check that the text is accessible, and record the deployment date. Leave the prompt untouched for the next run.

This is where the time box earns its keep. A [Mersel AI analysis](https://www.mersel.ai/blog/best-ai-visibility-tools-2026) describes the operational problem: teams can collect tracking data while failing to ship the pages and citations needed to act on it. I have seen the PPC version of that habit. Someone presents a beautiful search terms deck while the negatives remain unapplied. **If the log does not lead to an asset change, cut time from the log, not the change.**

## 6. When the constraint lifts: Hand off execution, not direction

**Cost: more than the three-hour DIY time box. Buy: capacity to work through a backlog instead of merely measuring it.**

The routine does not usually fail because the tracker stops running. It fails when the tracker finds more work than one person can ship. Say your log shows 20 structural gaps across the three engines and your calendar allows one page fix a week. Once your backlog of unaddressed prompt gaps exceeds 15 items, this is no longer an optimization loop. It is an inventory of work you do not have time to do.

Cross-engine gaps can deepen the problem. A page or technical issue affecting a live citation calls for different work from a missing presence in third-party discussions. One person spending an hour a week editing content cannot necessarily cover both, especially when search visibility feeds qualified pipeline or closed-won revenue. Do not buy more monitoring to disguise an execution bottleneck.

Keep the prompt list, commercial positioning, target margins, and decision about which gaps matter in-house. Hand off the repeated checking, page production, crawlability fixes, and citation work when those tasks no longer fit the calendar. A traditional agency retainer can put you back where you started: paying for human hours, a PDF deck, and a queue of fixes that moves too slowly. That is not the handoff I would choose.

An autonomous engine like [groas](https://groas.com/earned-search) is the alternative: specialized AI models handle ongoing prompt discovery and execution work, including intent-matched pages and technical fixes, while a named human strategist sets guardrails and remains accountable to revenue. The point is not to replace your judgment about your customers. It is to stop making your three available hours the limit on every fix.

Run the constrained version first. Keep the prompts fixed, ship one change at a time, and watch the backlog. **When the fixes outgrow the calendar, retire the manual execution—not the commercial discipline that told you what to fix.**

## Frequently Asked Questions

### How many prompts should I track to see if AI engines recommend my business?

Start with 20 prompts: 4 brand prompts, 10 category prompts, and 6 comparison prompts, all built from the actual searches buyers used in the last 90 days of Google Ads or Search Console data. Hold the wording steady week to week so changes in answers mean something. Twenty prompts across three engines already add up to 60 checks per run.

### What does it cost to monitor whether ChatGPT, Perplexity, and Google AI Overviews mention my brand?

The setup described here stays under $100 a month. Cited entry tiers include Otterly.AI at $29 a month for 15 prompts and Peec AI at $95 a month for up to 50 prompts; you need a tier that covers all 20 prompts. Skip enterprise suites priced at $295 to $499 a month, and keep checking all three engines, since only 11% of domains are cited by both ChatGPT and Perplexity.

### What is the difference between a mention and a citation in AI answers?

A mention means the AI engine names your brand in its answer, while a citation means the engine links to your site's URL as a source. The two are different outcomes, and your tracking sheet should record which one happened, including the linked URL when there is one.

### A competitor showed up in a ChatGPT answer instead of my brand. Should I panic?

No. AI answers vary across runs, and one benchmarking study found only 9.2% of cited URLs stayed identical across repeat runs of the same query. Wait for the same prompt to omit your brand on three consecutive weekly runs before treating it as a real problem worth fixing.

### What should I look for when a prompt repeatedly leaves my brand out of AI answers?

Sort the gap into two buckets: either no dedicated page on your site answers the specific question, or the page exists but buries the answer in narrative copy without direct answers, clear specifications, or a useful table. Also inspect the sources the engine cites, because third-party discussions, reviews, and forums get cited far more often than company blogs, and rewriting your homepage will not change those sources.

### How long should I wait after making a page change before checking AI engine answers again?

Publish the change, record the date, and submit the URL for reindexing in Google Search Console, then stop editing that asset for seven to ten days. Check the same prompt with the same wording on the same three engines on the second Tuesday after publication, and log any change from No to Yes as promising rather than a confirmed win based on one run.

### When should I stop running AI visibility work myself and hand it off?

When the backlog of unaddressed prompt gaps exceeds about 15 items, the routine is no longer an optimization loop but an inventory of work you cannot ship, and it is time to hand off execution. Keep the prompt list, positioning, target margins, and decisions about which gaps matter in-house, and hand off repeated checking, page production, crawlability fixes, and citation work.

## Related Posts

- [![Clay figure pours a wheelbarrow of articles into a wall funnel whose pipe is disconnected, so the pages pile on the floor beside an unused wrench.](https://groas.com/media/blog/e1a99963f159a4756e614472ee73400643c5900fa7bde718abe8fa563e8cdc77.png)](https://groas.com/post/can-chatgpt-even-read-your-site-a-45-min)
  ### [Can ChatGPT Read Your Site? A 45-Minute Check Before You Write More AEO Content](https://groas.com/post/can-chatgpt-even-read-your-site-a-45-min)
  [October 6, 2026 • 10 min read](https://groas.com/post/can-chatgpt-even-read-your-site-a-45-min)

  [Written by Alexander Perelman](https://groas.com/post/can-chatgpt-even-read-your-site-a-45-min)
- [![Cartoon of a marketer noting 'position two' after one pull of a slot machine whose reels are already spinning to a new result.](https://groas.com/media/blog/038bfefb4237637a1718ff66b379d5bdc42d57615f25797529aa5ef5682a4f7e.png)](https://groas.com/post/i-measured-ai-visibility-like-a-ppc-acco)
  ### [Six PPC Habits That Misread AI Visibility](https://groas.com/post/i-measured-ai-visibility-like-a-ppc-acco)
  [October 6, 2026 • 10 min read](https://groas.com/post/i-measured-ai-visibility-like-a-ppc-acco)

  [Written by David](https://groas.com/post/i-measured-ai-visibility-like-a-ppc-acco)
- [![A tiny worker on a ladder fixes up five model shopfronts while a giant dusty 'take a number' dispenser looms behind, its ticket queue ignored.](https://groas.com/media/blog/35c8ce6c06242ab08178711c5abe08dda0df51ec927a2dd6bc9770aa5f83bdbf.png)](https://groas.com/post/one-strategist-zero-developers-five-clie)
  ### [One Strategist, Zero Developers, Five Domains: A 30-Day AEO Launch Playbook](https://groas.com/post/one-strategist-zero-developers-five-clie)
  [October 6, 2026 • 10 min read](https://groas.com/post/one-strategist-zero-developers-five-clie)

  [Written by David](https://groas.com/post/one-strategist-zero-developers-five-clie)

## The Machines Already Run Search, You Should Probably Own One

[get started free](https://groas.typeform.com/to/xC1bQNUT)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/one-person-three-hours-a-week-under-100#webpage","url":"https://groas.com/post/one-person-three-hours-a-week-under-100","name":"One Person, Three Hours a Week, Under $100 a Month: An AI Visibility Playbook","description":"One person. Three hours a week. Less than $100 a month in software. That is enough to find out whether ChatGPT, Perplexity, and Google AI Overviews recommend…","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://groas.com/media/blog/3c161ea0021051c9cc8cbfd661c6f7cc716c54a8ef289c08a7c45337366c5759.png"},"breadcrumb":{"@id":"https://groas.com/post/one-person-three-hours-a-week-under-100#breadcrumb"},"dateModified":"2026-10-04T05:20:36.860Z","mainEntity":{"@id":"https://groas.com/post/one-person-three-hours-a-week-under-100#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/one-person-three-hours-a-week-under-100#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"AI For Google Ads","item":"https://groas.com/category/ai-for-google-ads"},{"@type":"ListItem","position":4,"name":"One Person, Three Hours a Week, Under $100 a Month: An AI Visibility Playbook","item":"https://groas.com/post/one-person-three-hours-a-week-under-100"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/one-person-three-hours-a-week-under-100#article","url":"https://groas.com/post/one-person-three-hours-a-week-under-100","mainEntityOfPage":{"@id":"https://groas.com/post/one-person-three-hours-a-week-under-100#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"One Person, Three Hours a Week, Under $100 a Month: An AI Visibility Playbook","description":"One person. Three hours a week. Less than $100 a month in software. That is enough to find out whether ChatGPT, Perplexity, and Google AI Overviews recommend your business, provided you treat buyer prompts like a PPC search terms report:…","datePublished":"2026-10-04T05:20:36.420Z","dateModified":"2026-10-04T05:20:36.860Z","image":{"@type":"ImageObject","url":"https://groas.com/media/blog/3c161ea0021051c9cc8cbfd661c6f7cc716c54a8ef289c08a7c45337366c5759.png"},"author":{"@id":"https://groas.com/author/david#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"articleSection":"AI For Google Ads","keywords":"Google Ads, AI For Google Ads","wordCount":2175,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/david#person","name":"David","jobTitle":"Founder & CEO @ groas","image":"https://groas.com/media/blog/363be9ad5255816654dac065d01bf1c6063110b400bb38cc199570267232027c.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"FAQPage","@id":"https://groas.com/post/one-person-three-hours-a-week-under-100#faq","mainEntity":[{"@type":"Question","name":"How many prompts should I track to see if AI engines recommend my business?","acceptedAnswer":{"@type":"Answer","text":"Start with 20 prompts: 4 brand prompts, 10 category prompts, and 6 comparison prompts, all built from the actual searches buyers used in the last 90 days of Google Ads or Search Console data. Hold the wording steady week to week so changes in answers mean something. Twenty prompts across three engines already add up to 60 checks per run."}},{"@type":"Question","name":"What does it cost to monitor whether ChatGPT, Perplexity, and Google AI Overviews mention my brand?","acceptedAnswer":{"@type":"Answer","text":"The setup described here stays under $100 a month. Cited entry tiers include Otterly.AI at $29 a month for 15 prompts and Peec AI at $95 a month for up to 50 prompts; you need a tier that covers all 20 prompts. Skip enterprise suites priced at $295 to $499 a month, and keep checking all three engines, since only 11% of domains are cited by both ChatGPT and Perplexity."}},{"@type":"Question","name":"What is the difference between a mention and a citation in AI answers?","acceptedAnswer":{"@type":"Answer","text":"A mention means the AI engine names your brand in its answer, while a citation means the engine links to your site's URL as a source. The two are different outcomes, and your tracking sheet should record which one happened, including the linked URL when there is one."}},{"@type":"Question","name":"A competitor showed up in a ChatGPT answer instead of my brand. Should I panic?","acceptedAnswer":{"@type":"Answer","text":"No. AI answers vary across runs, and one benchmarking study found only 9.2% of cited URLs stayed identical across repeat runs of the same query. Wait for the same prompt to omit your brand on three consecutive weekly runs before treating it as a real problem worth fixing."}},{"@type":"Question","name":"What should I look for when a prompt repeatedly leaves my brand out of AI answers?","acceptedAnswer":{"@type":"Answer","text":"Sort the gap into two buckets: either no dedicated page on your site answers the specific question, or the page exists but buries the answer in narrative copy without direct answers, clear specifications, or a useful table. Also inspect the sources the engine cites, because third-party discussions, reviews, and forums get cited far more often than company blogs, and rewriting your homepage will not change those sources."}},{"@type":"Question","name":"How long should I wait after making a page change before checking AI engine answers again?","acceptedAnswer":{"@type":"Answer","text":"Publish the change, record the date, and submit the URL for reindexing in Google Search Console, then stop editing that asset for seven to ten days. Check the same prompt with the same wording on the same three engines on the second Tuesday after publication, and log any change from No to Yes as promising rather than a confirmed win based on one run."}},{"@type":"Question","name":"When should I stop running AI visibility work myself and hand it off?","acceptedAnswer":{"@type":"Answer","text":"When the backlog of unaddressed prompt gaps exceeds about 15 items, the routine is no longer an optimization loop but an inventory of work you cannot ship, and it is time to hand off execution. Keep the prompt list, positioning, target margins, and decisions about which gaps matter in-house, and hand off repeated checking, page production, crawlability fixes, and citation work."}}]}]}
```