Search Meta ads by keyword, brand or slogan.

Cost guard: before each paid Spytrend call, run the free `estimate_cost` with the same tool and arguments, estimate the cost and stay within the user's stated budget; if the user has not stated a budget, tell them the estimate before the first paid call. If pricing consent is required, show the returned terms and stop until the customer explicitly agrees; never accept on the customer's behalf. If a required call does not fit the stated budget, make no paid call and return `Budget limit prevents required research`; omit an optional call that does not fit and disclose reduced coverage. When opening ads with `get_ad`, skip multilang ads (`is_multilang`) and never call `get_media` after `get_ad`. Retry once only when `retryable=true`, after `retry_after_seconds`. Read `params.ignored` after every search; an ignored required filter never counts as applied. If a search reports `pagination.total_status=restricted` (a free-plan archive window; `meta.plan_window_cutoff` names the boundary), tell the user the result covers only ads that started on or before that date, and never report a zero from it as "no such ads exist". Cost notices are not part of the final customer report.

Find active Meta ads launched in the last **30 completed UTC days** in **United States / US** whose returned ad text may contain the literal **Product** phrase **acne patches**.

Use Spytrend read-only. This is step 1 of 4. Do not produce the customer report yet.

Accept Search type only as Literal phrase, Product, Brand, Slogan or Offer; one supported ISO-2 country; and window 7, 14, 30 or 90 completed UTC days. Country is given as `Name / code`: use only the ISO-2 code after `/` in calls and only the country name before `/` in the answer. Trim whitespace only at the phrase edges. Require 2-80 characters. Preserve spelling, punctuation, case and language. Do not translate, correct, add synonyms or generate variants.

Set `date_to` to the last completed UTC date and calculate inclusive `date_from` for exactly the selected number of dates. Run exactly one `search_ads(source="meta", query=<literal phrase>, country=<ISO-2>, search_in="text", date_from=<date>, date_to=<date>, status_today="active", sort_by="most_reused_creative", sort_order="desc", dedupe=true, limit=10)` call. Retry once only for a retryable busy/timeout response and only after its returned wait interval. Do not retry invalid input or completed zero. Do not paginate or broaden.

Require query, country, text scope, launch dates, active status, reuse sort and deduplication to be applied without semantic substitution or ignored filters. Otherwise preserve a terminal `Unavailable` state for Step 4. A returned inactive row does not invalidate an otherwise successfully applied active-status route: mark only that row `ineligible - inactive` and continue with the remaining rows.

Treat the query as candidate discovery. Mark an eligible active row `visible exact` only when the complete phrase occurs contiguously and case-insensitively in returned title or body. Mark it `needs opening` when body text is explicitly omitted. Mark a returned inactive row `ineligible - inactive`. A partial-token or separate-word match is not exact. Deduplicate by ad ID and preserve server order.

Retain for the next step: validated inputs and calculated dates; route status; total rows; the ordered candidate IDs; `is_multilang` for every row; visible title/body or explicit omission marker; visible exact-match state; advertiser ID and name when returned; active status; launch and last-seen dates; format; CTA; landing URL/domain; creative ID; selection reuse; lifetime reuse; stable evidence URL when returned. Do not discard a candidate merely because an optional field is missing.

Reply only with a compact Step 1 checkpoint: route status, date window, row count and ordered candidate IDs with `visible exact`, `needs opening`, `not exact` or `ineligible - inactive`. Do not expose tokens or quotas. End with `Ready for Step 2.`

Cost guard: before each paid Spytrend call, run the free `estimate_cost` with the same tool and arguments, estimate the cost and stay within the user's stated budget; if the user has not stated a budget, tell them the estimate before the first paid call. If pricing consent is required, show the returned terms and stop until the customer explicitly agrees; never accept on the customer's behalf. If a required call does not fit the stated budget, make no paid call and return `Budget limit prevents required research`; omit an optional call that does not fit and disclose reduced coverage. When opening ads with `get_ad`, skip multilang ads (`is_multilang`) and never call `get_media` after `get_ad`. Retry once only when `retryable=true`, after `retry_after_seconds`. Read `params.ignored` after every search; an ignored required filter never counts as applied. If a search reports `pagination.total_status=restricted` (a free-plan archive window; `meta.plan_window_cutoff` names the boundary), tell the user the result covers only ads that started on or before that date, and never report a zero from it as "no such ads exist". Cost notices are not part of the final customer report.

Continue the retained **Product** search for **acne patches** in **United States / US**. This is step 2 of 4. Do not rerun discovery and do not produce the customer report yet.

If Step 1 preserved terminal `Unavailable` or completed zero, make no calls, preserve that state and reply with a compact checkpoint ending `Ready for Step 3.`

Otherwise open retained eligible active candidates sequentially in server order with `get_ad(id=<ad ID>)` until five ads verify or six candidates have been opened, whichever comes first. Skip rows marked `ineligible - inactive` or `is_multilang=true` without counting them toward the six-opening cap. Never substitute an ad outside that page. Retry each required detail call once only for a retryable busy/timeout response and only after its returned wait interval.

Retain an opened ad only when it is factually active and the complete phrase occurs contiguously and case-insensitively in opened title/body. For Literal phrase, exact occurrence is sufficient. For Product, the phrase must identify the product or product class. For Brand, it must identify a brand or branded product. For Slogan, it must function as a slogan or tagline rather than ordinary prose. For Offer, it must state the offer. Reject a literal match with the wrong role.

For every opened candidate retain: ad ID; pass/reject reason; advertiser ID and explicit name when returned; an 80-160 character context excerpt around the match; explicit offer; factual active status; launch and last-seen dates; format; CTA; exact landing URL/domain; creative ID; selection reuse; lifetime reuse; stable evidence URL. Preserve unavailable fields as unavailable. Preserve selection and lifetime reuse separately only when both are internally consistent. If selection reuse exceeds lifetime reuse or either scope is unclear, retain `Reuse comparison unavailable` plus the coverage inconsistency.

Reply only with a compact Step 2 checkpoint: number opened, verified IDs, rejected IDs with brief reasons, and whether the five-result or six-opening stop was reached. Do not expose tokens or quotas. End with `Ready for Step 3.`

Cost guard: before each paid Spytrend call, run the free `estimate_cost` with the same tool and arguments, estimate the cost and stay within the user's stated budget; if the user has not stated a budget, tell them the estimate before the first paid call. If pricing consent is required, show the returned terms and stop until the customer explicitly agrees; never accept on the customer's behalf. If a required call does not fit the stated budget, make no paid call and return `Budget limit prevents required research`; omit an optional call that does not fit and disclose reduced coverage. When opening ads with `get_ad`, skip multilang ads (`is_multilang`) and never call `get_media` after `get_ad`. Retry once only when `retryable=true`, after `retry_after_seconds`. Read `params.ignored` after every search; an ignored required filter never counts as applied. If a search reports `pagination.total_status=restricted` (a free-plan archive window; `meta.plan_window_cutoff` names the boundary), tell the user the result covers only ads that started on or before that date, and never report a zero from it as "no such ads exist". Cost notices are not part of the final customer report.

Continue the retained **Product** search for **acne patches** in **United States / US**. This is step 3 of 4. Do not rerun discovery or reopen ads. Do not produce the customer report yet.

For each distinct advertiser ID among retained verified ads, use the advertiser or page name returned with the opened ad, including a returned page or webmaster name, and call `get_advertiser(id=<exact advertiser_id>)` only when the opened ad returns no such name. Retry a required identity call once only for a retryable busy/timeout response and only after its returned wait interval. A failed identity lookup does not remove the verified ad; retain `Name unavailable`.

Group duplicate creative IDs. When creative ID is unavailable, group only rows with the same explicit advertiser identity, identical normalized full opened body and identical landing route. Do not merge merely similar copy, different routes or different advertisers. Retain the number of observed ads and all evidence IDs for each exact group (internal only). Do not infer an offer family.

Retain for Step 4 every verified result/group with advertiser name, context excerpt, observed ad count, offer, active status, launch and last-seen dates, format, CTA, exact landing route, reuse comparison or its explicit unavailability, creative ID when available, and stable evidence URLs. Also retain the Step 1 route status and dates, total discovered rows, number opened, and any required verification gap.

Reply only with a compact Step 3 checkpoint: resolved advertiser names, final group count, observed-ad count per group and any unresolved required evidence. Do not expose raw advertiser IDs, tokens or quotas. End with `Ready for Step 4.`

Using only retained evidence from Steps 1-3, write the final customer report for the **Product** phrase **acne patches** in **United States / US**. Make no new calls.

Use no table and stay under 300 words. Start with `# Active Meta Ads Containing "acne patches": <country name>`, then `**[N] verified ads · [M] advertisers · [Complete/Partial/Unavailable]**`. `N` is the number of verified observed ads before exact grouping, not the displayed group count.

Add `## Verified ads`. For each exact result/group show advertiser, concise context proving both the exact phrase and requested Search type role, observed-ad count when grouped, offer, factual active status and launch date, format, CTA, landing route, selection/lifetime reuse or `Reuse comparison unavailable`, and the stable spytrend.com evidence URL when returned (never a raw ad ID).

Add at most three `## Research coverage` bullets stating the launch window, bounded ten-row candidate sample, how many candidates were opened, exact-phrase verification and any material gap.

A completed discovery with successful required verification is `Complete`, including when optional creative ID, last-seen date, offer wording, advertiser name or reuse fields are unavailable. Completed discovery with no verified row is `Complete` zero; say none was verified in the bounded sample, not that no ad exists. A terminal discovery failure is `Unavailable`. Use `Partial` only when required phrase, active-status or landing evidence was lost for an otherwise retainable opened ad, and name the gap.

Do not expose service/tool names, calls, parameters, quotas, tokens, timing, retries, raw advertiser or ad IDs or rejected rows. Literal phrase reports textual occurrence without assigning a semantic role. Do not infer affiliation, authorization, infringement, counterfeit status, spend, clicks, conversions, revenue, performance, profitability or complete market coverage.

End exactly:

`Ad text, destination and creative-reuse data are observed research signals. They do not prove affiliation, authorization, spend, conversions, revenue or advertising performance.`