How to find a competitor's longest-running ads without a paid spy tool
No free ad library shows how well a competitor's ad performs. What you can see is which ads an advertiser has kept running, how many copies of one creative it runs, and which offers keep coming back. Those are weak clues, and each one can mislead.
Our read of one public advertiser's Meta ads on 2026-11-16 shows how. 186 of its ads had no stop date and a start date on or before 2026-09-30, but only 9 of them were seen in the latest scrape, and 124 ads shared a single start date. Use run length to pick ads worth studying, and check last-seen dates before you trust it.
Which free libraries show how long a competitor's ad has run?
Four official libraries cover most paid channels. Each one shows timing differently, and for ordinary ads none shows spend or conversions. TikTok is the exception: its Top Ads dashboard shows performance details for ads that TikTok selects.
| Library | What it shows about timing | What it does not show |
|---|---|---|
| Meta Ad Library | For an ad that is running now: the creative, the Page, and the date the ad started to run. Ordinary ads that did not reach the EU or UK are not archived after they stop.1 | Spend and impression ranges, except for ads about social issues, elections or politics. Reach and targeting detail only for those ads and for ads delivered to the EU or UK.1 |
| LinkedIn Ad Library | Ad preview, format, advertiser and payer. Dates the ad ran, only for ads targeted to the EU. Covers ads that ran after June 1, 2023 and keeps each ad for one year after its last impression.2 | Impressions and targeting outside EU-targeted ads. The help page lists no spend figure.2 |
| Google Ads Transparency Center | Advertiser name, location, payer and the ads the advertiser has served over a time period. You can filter by advertiser or website name, date and location.3 | The help page lists no start date for each ad, and no spend or results.3 |
| TikTok Creative Center, Top Ads | A collection TikTok calls high-performing ads, filterable by region, industry and objective. TikTok says the ads meet performance thresholds and the advertiser authorized them, and it shows performance details for each one.45 | Ads that did not meet TikTok's thresholds or whose advertiser did not authorize them. The thresholds are not stated in the pages we read.45 |
For the click-by-click steps in the Meta and LinkedIn libraries, use our guides: how to use the Meta Ad Library for competitor research and how to use the LinkedIn Ad Library for competitor research. This page covers only how to read run length and copies of an ad.
Which clues suggest an advertiser keeps running an ad?
Five, and all are weak. These are our readings, not measures any platform documents.
| Clue | What to read | Where | How it can mislead |
|---|---|---|---|
| Run length | Days between an ad's start date and today | Meta (running ads); LinkedIn (EU-targeted ads only)12 | An ad can stay on because nobody turned it off. Many ads can share one start date (see the example below). |
| Copies of one creative | The same copy or creative under several library IDs | Meta | Copies can be audience splits or tests. Adplexity, a competing spy tool, describes "N ads use this creative and text" as identical copies of one fixed ad and "This ad has multiple versions" as combinations Meta assembles from the advertiser's assets.6 We did not find either label in the Meta help page we read. |
| Same creative across placements | The platform icons on a card, or publisher_platforms in Hopkin |
Meta | The library does not say whether the advertiser chose the placements. |
| A repeated offer or angle | The same offer in newer ads after older ones stop | Any library you log over time | Offers repeat for seasons and campaigns. A repeat shows what the brand wants to say. It says nothing about how an ad did. |
| Refresh cadence | How many new ads appear each week, and whether old ones stay | A dated log you keep, or a tool that scrapes daily | A stopped ordinary Meta ad disappears, so only a dated log shows when it stopped.1 |
None of these clues shows results. Neither Meta's nor LinkedIn's help page lists clicks, conversions or return on spend for an ordinary ad, and an ad that runs for months may have a small budget behind it.12
A worked example: one advertiser's Meta ads, read on 2026-11-16
We ran this on Hopkin, our product, using public Ad Library data only. We picked Wispr Flow because it was already tracked in our Hopkin workspace. Nothing here comes from its ad account. The numbers are a dated snapshot of Hopkin's copy of the Meta Ad Library: the latest scrape finished 2026-11-15 at 22:53 UTC and we read it on 2026-11-16. They show ads seen, not spend or performance.
We used meta_ads_list_competitor_ads, which lists a tracked advertiser's ads from Hopkin's scraped corpus. Hopkin's docs say its filters work on these clocks: first_seen_since and first_seen_until use the platform's start date, or the date Hopkin first scraped the ad when there is none, and since means "last seen by a scrape on/after a date".7 Each ad row carries these fields:
| Field | Example value | What it tells you |
|---|---|---|
library_ad_id |
2157523015162535 | The ID in the public Ad Library link for the ad |
started_running_at |
2026-08-21 07:00 UTC | The start date Meta shows |
first_seen_at |
2026-10-08 | When Hopkin first scraped the ad. This is not a start date. |
last_seen_at |
2026-11-15 22:53 UTC | The latest scrape that saw the ad |
stopped_running_at, is_active |
null, true | No stop date recorded |
display_format, publisher_platforms |
video; Facebook and Instagram | Format and placements |
cards |
one entry per card | The actual copy of a dynamic ad, whose top-level copy can be a template such as {{product.name}} |
| Question | Result |
|---|---|
| Ads in the corpus with no stop date | 628 |
| Of those, seen in the 2026-11-15 scrape | 200 |
| No stop date and a start date on or before 2026-09-30 | 186 |
| Of those 186, seen in the 2026-11-15 scrape | 9 |
| Ads with 2026-09-28 as the first-appeared date (the rows we read show Meta's start date as 2026-09-28) | 124 |
Three things stand out.
First, a missing stop date does not prove an ad is still running. In the rows we read from the group of 186 (the first page of the list), last-seen dates ran from 2026-10-23 to 2026-11-12 for every ad except those 9. Those ads may have stopped, or the scrape may have missed them. We cannot tell which. If you filter only on active status and sort by start date, you will rank ads that may be gone. Adding since with the date of the latest scrape removes them.
Second, the 9 ads left are two kinds. One is a three-card dynamic ad aimed at sales reps, started 2026-06-19, which is 149 days before the 2026-11-15 scrape. The other eight are video ads, most about dictating while coding, all started on 2026-08-21 (86 days), all on Facebook and Instagram only. Seven of the eight link to the same domain with a different trybe= query parameter, which looks like per-ad tracking. That is our reading. Eight ads that start together do not give you eight separate run lengths to compare.
Third, a start date dates one listing. The idea behind it can be older. One piece of copy ("You talk. It types. That's the whole trick.", headline "Just speak. It writes.") appears under six library IDs. Two started on 2026-08-14 and four on 2026-09-28, each pointing to the same site with one of five different trybe= values. We compared the text, not the videos. If you read only the newest listing, you would date this message about six weeks too late.
What this example does not show
- Spend, impressions, clicks or conversions for any of these ads. The library does not give them for ordinary ads.1
- Whether any ad works. A start date and a last-seen date describe delivery, not results.
- Why 177 of the 186 were not seen in the latest scrape. They may have stopped, or the scrape may have missed them.
- Whether 2026-09-28 was a launch day or a re-upload.
- Whether the videos behind the six copies match. We compared text only.
- Whether Hopkin's copy of the library is complete for this advertiser. We did not check it against the live Ad Library. Our attempt to open one ad's live library page from our tools failed, so the start dates are as Hopkin recorded them.
How do I do this for free, step by step?
You need a spreadsheet and a fixed day each week.
- Meta: search each competitor in the Ad Library, following our Meta guide. For every ad you care about, record the library ID, the start date, the copy, the destination URL and the date you checked.
- Recheck on the same day every week. An ordinary Meta ad drops out of the library when it stops, so an ad that is missing this week has stopped.1 Your dated log is the only record of that.
- Sort your sheet by copy text, not by library ID. Copies of one creative then sit together, and you can take the earliest start date as the age of the idea.
- Mark any start date shared by many ads as a batch and do not rank those ads against each other.
- LinkedIn: run dates appear only for EU-targeted ads, so write down the date you first saw each ad, as in our LinkedIn guide.2
- Google: use the Ads Transparency Center's date and location filters to compare periods. Its help page lists no per-ad start date, so log first-seen dates yourself.3
- TikTok: open Creative Center, Top Ads, and filter by region, industry and objective. Treat it as a gallery of formats TikTok selected, since advertisers must authorize their ads to appear there.45
For one to three competitors this is enough. Our weekly competitor ad report template shows how to report the results.
Can Claude or ChatGPT sort the ads for me?
Yes, if the assistant can read ad-library data. Hopkin's Meta and LinkedIn tools do that for advertisers you track, and Hopkin refreshes them daily.7 Hopkin does not track competitor ads on Google or TikTok. For those, use the libraries by hand as above.
Reading a tracked advertiser's ads is free on Hopkin's pricing page, a search costs 5 credits, and tracking a competitor is listed at 200 credits a month. Starter is $100 a month for 1,500 credits, and the 30-day trial gives 1,000 credits with no card (pricing read 2026-11-16).8
Below is a prompt for sorting delivery data. It is not a performance verdict, and it tells the assistant not to rank ads by results. We ran the underlying tool calls ourselves for the example above. We did not test this prompt text in Claude or ChatGPT.
You have Hopkin's Meta ad library tools. Use public ad library data only. 1. Call meta_ads_list_tracked_competitors and find <competitor>. If it is not tracked, ask me before tracking it. 2. Call meta_ads_list_competitor_ads for that advertiser with active_only true and since set to the date of the latest scrape. Page through every result. 3. Group ads that share the same primary_text or the same card text. For each group, report: number of library IDs, earliest and latest started_running_at, platforms, formats and destination domains. 4. For each group, count the days from the earliest started_running_at to the latest last_seen_at. If 10 or more ads share one start date, label them a batch launch and do not rank them by run length. 5. In a separate list, show ads flagged active whose last_seen_at is older than the latest scrape, labelled "may have stopped". 6. Return one table sorted by days running with these columns: creative group, copies, platforms, first start date, days running, caveat. 7. Do not say any ad performs well, wins or works. You have no spend, impressions or results. End with the three things this data cannot tell me.
On LinkedIn, Hopkin's linkedin_ads_list_competitor_ads returns ads with run dates that can be empty until a scrape has read each ad's detail page, so expect fewer dates there.9
How this page was made
An AI agent drafted this page. It read the library help pages for Meta, LinkedIn, Google and TikTok, Hopkin's docs and pricing page, and one competing vendor's article, all on 2026-11-16. It also ran Hopkin's Meta ad library tools on one public advertiser on 2026-11-16, as described above. It did not run Hopkin on LinkedIn, Google or TikTok data, did not test the prompt in Claude or ChatGPT, did not compare Hopkin's data with the live Meta Ad Library, and did not test any paid spy tool. Hopkin is our product, and its tools are the only ones we ran. No person reviewed this page before publication.
Frequently asked questions
How do I find a competitor's best-performing ads without paying for a spy tool?
You cannot see performance for ordinary ads in any of the free libraries. You can find ads that an advertiser has kept running, copies of one creative and repeated offers, and treat them as ads worth studying. TikTok's Top Ads dashboard is the one official library that selects ads by performance, by TikTok's account of it, and it covers only ads whose advertisers authorized them.
How can I tell how long a competitor has been running an ad?
On Meta, an ad that is running shows the date it started, so count from that date to today. On LinkedIn, run dates appear only for ads targeted to the EU. Google's help page lists no per-ad start date, so keep a dated log of what you see. Whatever the platform, check that the ad is still there, and find the earliest start date among copies of the same creative.
Is a long-running ad a winning ad?
It is a weak clue. An ad can run for a long time because it works, because nobody turned it off, or because little money is behind it, and the libraries do not say which. In the snapshot above, 124 ads shared one start date, so run length ranked ads that were launched together.
Does TikTok Creative Center show top-performing ads?
TikTok describes Top Ads as "real TikTok ads that meet performance thresholds and advertiser authorization to appear on Creative Center". The pages we read do not state the thresholds, and ads that advertisers did not authorize are not included.
Can the Google Ads Transparency Center show how long an ad has run?
Google's help page says you can search advertisers and see the ads they have served over a time period, filtering by date and location. It lists no start date for each ad. Log the date you first see each ad.
Can Claude or ChatGPT sort a competitor's ads by how long they have run?
Yes, when it can read ad-library data. With Hopkin connected, it can list a tracked advertiser's Meta ads with start and last-seen dates and group copies of one creative. The prompt above does this and is built to avoid calling any ad a winner.
Sources
All checked .
- About the Meta Ad Library, Meta Business Help Center. Supports: what the library shows by type of ad; start date for running ads; no archive for ordinary ads; spend, impression ranges, reach and targeting by type of ad. Checked .
- LinkedIn's Ad Library, LinkedIn Marketing Solutions Help. Supports: fields shown; dates, impressions and targeting for EU-targeted ads; coverage since June 1, 2023; one-year retention. Checked .
- Ads transparency, Google Advertising Policies Help. Supports: what the Ads Transparency Center shows and the date and location filters. Checked .
- Creative Center, Top Ads one-pager, TikTok. Supports: Top Ads meet performance thresholds and advertiser authorization; performance details shown per ad. Checked .
- About Top Ads, TikTok Ads Help Center (last updated May 2025). Supports: Top Ads are filterable by region, industry and objective; all featured ads are authorized by advertisers. Checked .
- How to Use the Meta Ad Library in 2026, Adplexity (a competing spy tool). Supports: its description of the "multiple versions" and "N ads use this creative and text" labels. Third-party, not Meta documentation. Checked .
- meta_ads_list_competitor_ads, Hopkin docs. Supports: filters, the first_seen and since clocks, daily scrape of tracked advertisers. Checked .
- Pricing, Hopkin. Supports: Starter, credits, ad library search, competitor tracking, free reads of tracked competitors' ads, trial. Checked .
- How to use the LinkedIn Ad Library for competitor research, Hopkin. Supports: Hopkin's LinkedIn run dates can be empty until a scrape has read each ad's detail page. Checked .
- Hopkin tool output: meta_ads_list_tracked_competitors and meta_ads_list_competitor_ads for the Wispr Flow advertiser (Page ID 110864474787474), run on a corpus synced 22:53 UTC. Supports Tables 3 and 4 and the three findings. Public Ad Library data only.