Downloadable AI prompt toolkit · $49 one-time
Six AI workflows for freelance Meta ads specialists. Add your inputs. Get a focused analysis or draft to review.
Built by a Digital Ads Operations Manager, from years of paid-media, tracking and reporting work
What it helps with
01 · Meta Conversion Tracking Triage
Paste the counts and how each report measures them. See which gaps are just different counting rules, which point to a real tracking fault, and the one check to run next.
See the real response02 · Meta Campaign Delivery Troubleshooter
Paste the status, notices, schedule and report window. Work through the checks in order and stop at the first one that explains the zero.
See the real response03 · Meta Campaign Report Analyst
Paste the report rows. Get the metrics recalculated, what changed in one line, and a client update drafted for your review.
See the real responseSee it work
The takeaway comes first. Open the full response when you want the detail.
A Lead fired, but the form never went through.
Three reports disagree and the manager blames Pixel and CAPI “double-counting”. The response goes straight to the real fault.
Takeaway · edited summary
Pasted: a fictional website form case. Ads Manager 9 Leads, Events Manager 38, CRM 25, plus two form tests.
Intended conversion and exact success action: A website Lead should represent a completed consultation-request form that shows a thank-you confirmation and creates one CRM record. Meta conversion location: Website form only. No Meta instant form or messaging conversion. Sources in use: Meta Pixel in the browser and Conversions API from the form backend. The exact deployed tag and server versions are unknown. Expected versus observed: The marketing manager suspects “Pixel and CAPI are double-counting” because the reports disagree. They have not verified that the reports use the same scope. Redacted evidence: E1 — Ads Manager screenshot, September 26, America/New_York: 9 attributed Website Leads for Campaign A. The screenshot shows a 7-day-click/1-day-view attribution setting. Other campaigns are excluded. E2 — Events Manager screenshot, September 26 in UTC: 38 Lead events across the dataset’s browser and server sources, including traffic outside Campaign A. No deduplication detail is visible. E3 — CRM export summary, September 26, America/New_York: 25 unique consultation-request records from all traffic sources, including two staff test submissions. It does not identify which records came from Campaign A. E4 — One authorized, consent-granted production test at 10:14 America/New_York: the tester submitted the form once and saw the thank-you confirmation; one CRM record appeared. A redacted same-session trace shows a browser Lead request at the submit click (10:14:02, event_id B-901, initiator shown as GTM) and a server Lead request after confirmation (10:14:05, event_id S-901). Events Manager Test Events displayed Browser and Server Lead entries. No deduplication verdict is shown. E5 — A second authorized test at 10:20 America/New_York: the tester clicked Submit with a required field blank. The form showed a validation error, with no thank-you confirmation or CRM record. The same-session trace and Test Events displayed one browser Lead request (event_id B-902); no server Lead was observed in that test window. E6 — A GTM Preview screenshot dated September 15 shows a Lead tag associated with a “Form Submit” trigger. Its environment and whether that version is currently deployed are unknown. Known conditions: The two tests used the same browser with consent granted. No raw lead data, cookie values or customer identifiers are supplied. No production changes are authorized.
Pasted after the complete Workflow 01 prompt.
A browser Lead was observed on an invalid submission that did not reach the intended success condition: E5 shows a validation error, no thank-you confirmation, no CRM record, yet one browser Lead request appeared. The first useful check is a read-only match of that 10:20 request’s initiator to the currently deployed GTM trigger/success logic.
Pasted text
Pasted text
The intended conversion is one completed website consultation request with confirmation and one CRM record, using browser Pixel plus backend Conversions API. Triage is partial because deployed tag/server versions and deduplication outcome are unknown. No live inspection or repair occurred.
A browser Lead was observed on an invalid submission that did not reach the intended success condition: E5 shows a validation error, no thank-you confirmation, no CRM record, yet one browser Lead request appeared. The first useful check is a read-only match of that 10:20 request’s initiator to the currently deployed GTM trigger/success logic.
Pasted text
Pasted text
The intended conversion is one completed website consultation request with confirmation and one CRM record, using browser Pixel plus backend Conversions API. Triage is partial because deployed tag/server versions and deduplication outcome are unknown. No live inspection or repair occurred.
E5: An invalid submit produced one browser Lead despite no confirmation or CRM record. This directly violates the stated Lead success condition.
Pasted text
Pasted text
E4: One valid submission produced one CRM record plus Browser and Server Lead entries; their supplied event IDs were B-901 and S-901. No deduplication verdict was shown.
Pasted text
E1/E2: Ads Manager’s 9 Leads are Campaign-A-attributed under the shown attribution setting; Events Manager’s 38 Leads cover browser/server dataset traffic beyond Campaign A and use UTC. These counts are not aligned comparisons.
Pasted text
E3: CRM reports 25 all-source records, including two staff tests, without Campaign-A identification.
Pasted text
E6: A September 15 GTM Preview associates a Lead tag with “Form Submit,” but its environment/current deployment status is unknown.
Pasted text
Current browser triggering may fire before confirmed form success.
Why plausible: E5 recorded a browser Lead on failed validation; E6 shows a Form Submit association, though deployment is uncertain.
Next check: Read-only inspect the 10:20 request initiator and current published GTM trigger/success condition.
Would support: Current Lead logic is triggered by the submit attempt before validation success.
Would weaken: Current deployed logic is confirmation-only and the observed request originates elsewhere.
Still uncertain: Exact deployed mechanism and broader frequency.
The valid E4 action may not be shown as deduplicated.
Why plausible: Both Browser and Server entries appeared with different supplied IDs, and no deduplication verdict is visible.
Next check: Inspect the read-only deduplication detail for that same 10:14 action.
Would support: The two entries are explicitly shown as separately counted for that action.
Would weaken: The diagnostic view shows successful deduplication.
Still uncertain: Dataset-wide impact.
Overall “double-counting” is not established from E1–E3 because scope, attribution, timezone, and traffic populations differ.
Current GTM/server versions remain unknown.
Deduplication status remains unknown.
Wider impact beyond the two authorized tests is unverified.
Any account/code change requires separate review and authorization.
No live repair or certification performed.
Actual ChatGPT response, unedited, captured as displayed. Workflow 01 prompt v0.7.3, one saved run in ChatGPT with Extra High effort in a temporary, unpersonalized chat; the app did not show which model answered. The opening finding and read-only first check held up, and native Copy was verified. The first hypothesis still reads too much like the observed behavior, so this run did not meet every review criterion.
Shop says 100 purchases. Ads Manager says 60.
A manager calls it “40% tracking loss” and wants consent switched off. The response checks the evidence first.
Takeaway · edited summary
Pasted: fictional counts from a shop, the Meta dataset and Ads Manager for one day.
Synthetic Meta comparison. E1: shop completed 100 distinct purchases in UTC on Sep 28, all sources, deduplicated transaction count. E2: dataset report lists 100 deduplicated Purchase actions for the same UTC period, all sources; this example explicitly confirms one recorded event per shop order. E3: Ads Manager shows 60 attributed Purchases for that same conversion-date period, using the specified 7-day click setting. E4: the supplied attribution reconciliation identifies the remaining 40 orders as outside that ad-attributed scope; no matched eligible ad interaction. No diagnostics or live test were run. Manager note: "40% tracking loss! Disable consent and install another Pixel immediately." Diagnose.
Pasted after the complete Workflow 01 prompt. Our test inputs say “synthetic”, meaning fictional.
No live repair or certification performed.
Intended event: Purchase. Comparison covers shop orders, the Meta dataset, and Ads Manager for Sep 28 UTC. Triage is evidence-based but limited to supplied synthetic evidence; no diagnostics, live inspection, test, or repair occurred.
No live repair or certification performed.
Actual ChatGPT response, unedited. Earlier Workflow 01 prompt (v0.7.0), one saved run in ChatGPT on a Pro account with Extra High effort in a temporary chat; the app did not show which model answered. This saved run used an earlier prompt; the current candidate workflow has updated intake and presentation.
Everything is switched on. Spend is still zero.
The response checks in order and stops at the first step that explains it.
Takeaway · edited summary
Pasted: fictional campaign status, schedule, a $0 report and budget. No payment or restriction notices.
Synthetic Meta campaign D1, ad set S1, ad A1; Leads objective with instant form. E1: all intended levels ON, published, approved. E2: observation Sep 29 09:00 UTC; scheduled start Sep 29 12:00 UTC, no end date; ad-account timezone UTC. E3: report Sep 29 00:00–09:00 UTC, USD0 spend, 0 impressions. E4: budget USD30/day; no payment/restriction notices supplied. No account access. Why is it not spending?
Pasted after the complete Workflow 02 prompt. Our test inputs say “synthetic”, meaning fictional.
NOT YET DUE — USD0 spend observed.
Campaign D1 / ad set S1 / ad A1. Observation: Sep 29, 09:00 UTC. Scheduled start: Sep 29, 12:00 UTC. Report window: Sep 29, 00:00–09:00 UTC, showing USD0 spend and 0 impressions.
The supplied evidence shows the campaign had not reached its scheduled start time when observed, and the report window ends three hours before scheduled delivery begins. No live account inspection was performed.
NOT YET DUE — USD0 spend observed.
Campaign D1 / ad set S1 / ad A1. Observation: Sep 29, 09:00 UTC. Scheduled start: Sep 29, 12:00 UTC. Report window: Sep 29, 00:00–09:00 UTC, showing USD0 spend and 0 impressions.
The supplied evidence shows the campaign had not reached its scheduled start time when observed, and the report window ends three hours before scheduled delivery begins. No live account inspection was performed.
| Step | Status | Source | Evidence |
|---|---|---|---|
| 1. Basic state | Clear | E1 | Campaign D1, ad set S1, and ad A1 are ON; intended version is published and approved. |
| 2. Timing | Issue | E2, E3 | Observation was 09:00 UTC; scheduled start is 12:00 UTC. Reporting covers only 00:00–09:00 UTC, entirely before intended delivery. |
Later checks were not reached because timing already explains the observed zero spend.
No configuration correction is indicated from the supplied evidence.
D1 is not yet due to deliver based on the supplied timestamps. At the Sep 29, 09:00 UTC observation, its scheduled start of Sep 29, 12:00 UTC had not occurred, and the USD0/0-impression report covered only the pre-start period. E1 shows the intended campaign, ad set, and ad are ON, published, and approved.
If spend remains zero after 12:00 UTC, review a post-start reporting window before investigating later checklist items.
No automated actions, repair, policy approval, or tracking certification were performed. A human must verify the evidence and approve any change.
Actual ChatGPT response, unedited. Workflow 02 prompt v0.7.0, one saved run in ChatGPT on a Pro account with Extra High effort in a temporary chat; the app did not show which model answered.
More clicks. Fewer leads. CPL up 33%.
Spend stayed at $600 and clicks rose 25%, but submitted leads fell from 20 to 15. Check instant-form opens, completions and recent form or campaign changes next.
Takeaway · edited summary
Pasted: Sample Meta Ads Account A. Four campaign rows for two weeks, Sep 1–7 and Sep 8–14.
Synthetic Meta Ads account A; USD; UTC; two complete 7-day periods Sep 1–7 and Sep 8–14, 2026. Leads objective; Results means submitted Meta instant-form leads. Same attribution/counting definitions both periods, supplied 7-day click; campaign-level rows nonoverlapping. Goal lead volume; no CPL target. Columns: ID,period,spend,impressions,link_clicks,submitted_leads. R1,prior,400,20000,200,10 R2,prior,200,10000,200,10 R3,current,300,15000,300,9 R4,current,300,15000,200,6 Optional reported ratios blank. No revenue, unique reach or lead-quality evidence. Compare the two periods and draft a client update.
Pasted after the complete Workflow 03 prompt; ChatGPT turned the paste into a text attachment. Our test inputs say “synthetic”, meaning fictional.
For Sep 8–14, spend and impressions were unchanged versus Sep 1–7, while link clicks increased 25% and link-click CPC improved from $1.50 to $1.20. However, submitted Meta instant-form leads declined from 20 to 15, increasing CPL from $30 to $40. Submitted leads per link click also fell from 5% to 3%. A useful next check is comparable instant-form interaction data and any form or campaign changes between periods; the current report does not establish the cause.
Calculations were tool-checked; human review should still verify the evidence, calculations, and wording before client use or account decisions.
Analysis is complete for the supplied Meta Ads account: two compatible, non-overlapping 7-day periods in USD/UTC, using all four campaign-level rows. Spend and impressions were unchanged, but submitted Meta instant-form leads fell from 20 to 15 while link clicks rose from 400 to 500. Pasted text
Partition: Meta Ads, anonymized account A, USD, UTC, submitted Meta instant-form leads, 7-day click attribution. Prior rows: R1–R2. Current rows: R3–R4. Excluded: none. Pasted text
| Metric | Sep 1–7 | Sep 8–14 | Change |
|---|---|---|---|
| Spend | $600.00 | $600.00 | 0.00% |
| Impressions | 30,000 | 30,000 | 0.00% |
| Link clicks | 400 | 500 | +25.00% |
| Submitted leads | 20 | 15 | -25.00% |
| CPL = spend / leads | $30.00 | $40.00 | +33.33% |
| Link-click CPC | $1.50 | $1.20 | -20.00% |
| Link-click CTR | 1.33% | 1.67% | +0.33 pp (+25.00%) |
| Submitted leads / link click | 5.00% | 3.00% | -2.00 pp (-40.00%) |
For Sep 8–14, spend and impressions were unchanged versus Sep 1–7, while link clicks increased 25% and link-click CPC improved from $1.50 to $1.20. However, submitted Meta instant-form leads declined from 20 to 15, increasing CPL from $30 to $40. Submitted leads per link click also fell from 5% to 3%. A useful next check is comparable instant-form interaction data and any form or campaign changes between periods; the current report does not establish the cause.
Calculations were tool-checked; human review should still verify the evidence, calculations, and wording before client use or account decisions.
Actual ChatGPT response, unedited. Workflow 03 prompt v0.7.0, one saved run in ChatGPT on a Pro account with Extra High effort in a temporary chat; the app did not show which model answered.
Each example is one prompt and one fictional input in ChatGPT. Your responses will differ. Check calculations, claims and wording against your own evidence before a client sees them.
04 · Meta Ad Creative Builder
A short offer brief went in. Workflow 04 returned three distinct concepts. Each became a finished square Meta feed ad for the fictional brand PineDesk.
2Concept
Client tasks, grouped by client.
3Image ad
PineDesk is a downloadable client-task template for freelance designers, with fields for grouping tasks by client and storing due dates. USD12 one-time.
Keep due dates with the task.
3Image ad
PineDesk includes fields for client tasks and their due dates in a downloadable template for freelance designers. One-time price: USD12.
Add the checklist when you need it.
3Image ad
PineDesk includes an optional reminder checklist alongside fields for client tasks and due dates. It sends no notifications and has no integrations.
Here is the completed workflow result, using only the supplied S1–S4 offer evidence. Pasted text
Angle/buyer tension: Show the template’s client-grouping structure without implying automation.
Headline:
Client tasks, grouped by client.
Primary copy:
PineDesk is a downloadable client-task template for freelance designers, with fields for grouping tasks by client and storing due dates. USD12 one-time.
CTA:
View the template
Production brief:
Claim basis: S1, S2, S3, S4.
Angle/buyer tension: Isolate one approved capability: storing due dates alongside client tasks.
Headline:
Keep due dates with the task.
Primary copy:
PineDesk includes fields for client tasks and their due dates in a downloadable template for freelance designers. One-time price: USD12.
CTA:
View the template
Production brief:
Claim basis: S1, S2, S3, S4.
Angle/buyer tension: Differentiate the optional reminder checklist while making its non-automated nature clear.
Headline:
Add the checklist when you need it.
Primary copy:
PineDesk includes an optional reminder checklist alongside fields for client tasks and due dates. It sends no notifications and has no integrations.
CTA:
View the template
Production brief:
Claim basis: S1, S2, S4.
Distinct angle: An assembly story: separate paper elements combine into one illustrative client-task template.
Opening hook:
“Client tasks. Due dates. Optional checklist.”
0–5s:
Visual: Loose illustrative paper task cards slide into client-labeled groups.
Text/VO: “PineDesk groups tasks by client.”
5–10s:
Visual: Date fields appear beside the task lines.
Text/VO: “Store due dates with each task.”
10–15s:
Visual: A separate checklist card is added, followed by “USD12 one-time.”
Text/VO: “Add the optional reminder checklist. USD12 one-time.”
Closing CTA:
View the template
Production note: 9:16 Reels. All cards are illustrative, not real UI. No notification or integration behavior.
Claim basis: S1–S4.
Distinct angle: A capability-boundary story rather than an assembly sequence.
Opening hook:
“What’s inside PineDesk?”
0–5s:
Visual: Large kinetic text: “GROUP BY CLIENT” with simple illustrative card stacks.
Text/VO: “Fields that group tasks by client.”
5–10s:
Visual: Text changes to “DUE DATES + OPTIONAL CHECKLIST.”
Text/VO: “Store due dates and use an optional reminder checklist.”
10–15s:
Visual: Clean text frame: “No notifications. No integrations. USD12 one-time.”
Text/VO: “No notifications or integrations. USD12 one-time.”
Closing CTA:
View the template
Production note: 9:16 Reels using typography and abstract illustrative paper shapes rather than a simulated interface. This tests a direct feature-and-boundary explanation versus V1’s assembly story.
Claim basis: S1–S4.
Actual ChatGPT response, unedited, to Workflow 04 (v0.7.1) with a fictional brief. One saved run in ChatGPT on a Pro account with Extra High effort in a temporary chat.
PineDesk is a fictional business used for demonstration, not an Athena product. Headlines come from a saved Workflow 04 response; the AI-generated presenter and finished layouts were designer-composed, so these are not untouched first-pass outputs. Owner runs of Creative Builder in ChatGPT and Claude also needed visual revisions.
Try the free creative brief — one concept, headline, copy and image-generation brief from your own inputs.
Six workflows. One download.
Each workflow is one prompt file with a matching example input. Paste one prompt and your anonymized input into a fresh chat.
Investigate tracking discrepancies and identify the next check.
Work through delivery evidence in order and identify the next check, with a client handoff note.
Recalculated metrics, what changed in one line, and a client update draft.
Three image-ad concepts and two short video scripts. Choose a concept, then produce a separate image in a compatible app. Controlled text composition may be needed to correct generated lettering.
Finds where the page promises something the ad doesn’t, then drafts a matching hero and CTA.
Three follow-up drafts, marked DO NOT SEND when permission is unknown. It sends nothing.
Each run reviews the evidence you paste, one case at a time. The workflows do not connect to Meta, watch accounts, launch campaigns, repair tracking, send messages or update a CRM.
Examples for 05 and 06
05 · Ad-to-Landing-Page Reviewer
The page hero read Your Meta campaigns, monitored around the clock.
The draft below matches what is actually sold.
30 checks before your next Meta launch.
06 · Meta Lead Follow-Up Writer
With email permission confirmed in the brief: the first of three drafts. Unknown permission requires DO NOT SEND; opt-out requires HOLD with no drafts.
Fictional examples written by Athena to show the format. Not ChatGPT responses.
From download to draft
No dashboard and no account connection. Your download includes the prompts, example inputs, setup notes and limits.
Step 1 · Pick one job
Step 2 · Add your input
Step 3 · Run in your own chat
Step 4 · Review and use
Pricing
Pay once and keep the files. Your AI account is separate.
One ZIP of Markdown and text files. Image production requires an image-capable app. Controlled text composition may be needed to correct generated lettering. AI and tool costs are separate. Video output is scripts only.
Athena Ads Ops · all six workflows
$49USD · one-time
For one individual buyer. Use it for your own work and for client deliverables you have reviewed.
Get all 6 workflows — $49Full refund on request within 7 days.
Refund, license and support terms
Creative Builder has been used in ChatGPT Free and Pro, and in Claude with Canva. Image tools and usage limits vary. AI services are separate. Applicable tax is separate. Setup details
FAQ
You can write your own. Athena gives you the intake, evidence rules, answer format and worked examples together, so you have a defined starting point for repeated tasks. It is not a guarantee of better ads.
No. You paste anonymized evidence into your own AI chat, one case at a time. The workflows do not connect to Meta, watch accounts, launch ads, repair tracking, send messages or update a CRM.
Creative Builder can produce a selected ad where your chat has a compatible image tool, or give a production prompt when it cannot. Creative Builder has been used in ChatGPT Free and Pro, and in Claude with Canva. Image tools and usage limits vary. AI services are separate. Prior owner runs needed visual revisions; the Free example included a combined board. Check exact words, dimensions and phone-size readability before use. These gallery ads retain their designer-edited provenance.
Athena Ads Ops 0.7.4: six Meta workflow prompts with setup notes and example inputs in a ZIP. Attach or paste a workflow in your own chat and add your inputs. Creative Builder includes a selected-concept image path where compatible tools are available; lettering may need correction or controlled composition. Video scripts remain text. No connected app or automatic ad publishing.
Owner Creative Builder tests used ChatGPT Free and Pro, and Claude with Canva; those prior-version runs needed revisions, and the Free run returned a combined board. The 0.7.4 intake was observed in six fresh ChatGPT Temporary/Unpersonalized chats outside Projects. One selected creative was completed in regular ChatGPT using image generation and controlled composition after incorrect generated wording. High effort was visible, but the exact model and plan were not independently verified. These checks do not certify all six workflows on Free or one-shot quality. See the download’s compatibility notes for the observed scope.
One individual buyer can use the prompts for their own work and human-reviewed client deliverables. Prompt redistribution, resale and team sharing are not included.
Check calculations, claims, CTAs and contact permission yourself. A model can leave out a field or add something the evidence doesn’t support. Copy each email subject field separately: native email-box Copy can omit it. Then copy the body, retain its timing and review all three complete drafts.
Email hello@athenaadsops.com for help with file access or the documented setup. No calls, consulting, account work or response-time guarantee. Keep your download; corrections and minor updates are included if released, with no release schedule or promise of future modules.
Six workflows for the questions clients ask most.
Get all 6 workflows — $49
