Most guides to spying on competitor ads are a guide to the Meta Ad Library with a paragraph bolted on about Google. That's backwards for a lot of accounts, and it skips the single most useful competitor report in the industry — one that's free, first-party, and sitting unopened in your own Google Ads account.
This playbook covers both platforms properly: what's genuinely visible on each, which signals mean something, which are noise, and the weekly routine that turns all of it into briefs instead of a folder of screenshots.
First: what you can actually see
Being precise about this saves a lot of wasted effort.
| Meta | ||
|---|---|---|
| Every live creative | ✓ Public library | ✓ Transparency Center |
| How long an ad has run | ✓ Start date shown | Partial — dates shown, less granular |
| How many copies of an ad run | ✓ Visible | Harder |
| Landing pages | ✓ | ✓ |
| Search terms they bid on | ✗ | Inferred via keyword tools |
| Whether you compete with them at all | ✗ | ✓ Auction Insights, free |
| Their share of the auction | ✗ | ✓ Auction Insights, free |
| Spend | ✗ | ✗ |
| ROAS, CPA, conversion rate | ✗ | ✗ |
Nobody can see spend or return. Any tool claiming to is modelling it, and the model is usually impressions × a guessed CPM. Treat those numbers as decoration.
The two rows in bold are why the Google half of this matters. On Meta you can see everything except whether you're actually in the same auction. On Google you can see exactly that.
Part 1 — The Meta playbook
Where to look
The Meta Ad Library is the primary source and it's free: search any advertiser, filter by country and date, see every live ad. Everything else in this category reads from it and adds memory, tagging and alerts.
The four signals that matter
1. Longevity. How long an ad has run. The only honest public proxy for performance — nobody keeps paying to serve a loser. Anything past 60 days is a verdict.
2. Copy count. How many near-identical versions are live. One long-running ad can be luck; the same hook rebuilt and re-shipped is a decision.
Perfora is a clean example. Its plaque-on-braces demo runs as two separate ads, 32 and 29 days, with the same visual in both — the brand rebuilt it rather than letting one expire. Alongside it:
The whitening-strips ad, on a completely different promise, is still running at 59 days. Two signals from one brand in one glance: a demo they believe in enough to duplicate, and a separate concept outliving both copies of it.
3. The variant fan-out. When a brand suddenly ships six versions of one concept, they've found something. The shape of the fan-out tells you what they're testing — six hooks on one visual is a messaging test; one hook across six visuals is a format test.
4. Offer changes. The most actionable signal and the one almost nobody logs, because it requires reading the copy rather than looking at the picture.
V-Guard's English rooftop-solar ad names a government subsidy of up to ₹78,000. It ran 73 days.
The Hindi execution of the same concept names up to ₹1,08,000. It ran 70 days, in parallel.
Same brand, same offer structure, same quarter, two different subsidy figures — a regional or scheme variation, stated in rupees, on the face of the creative. A competitor tracker that only stores images would never surface it. A screenshot folder definitely won't. This is why the instruction is read the copy, not look at the ad.
What to log for every Meta ad
Brand, hook type, offer and its named job, format, first seen, still live, number of near-identical copies, landing page URL, and one line on what you'd steal. The last field is the one everyone skips and later wishes they had.
Part 2 — The Google playbook
This is where most competitor research stops being done properly, so it's worth doing in order.
1. Google Ads Transparency Center — what they're saying
Google's own library shows the ads an advertiser is running, by format and region, including search, display, YouTube and shopping. It's the Meta Ad Library equivalent and the right first stop.
What it gives you: actual headlines and descriptions, the creative, the regions, and roughly how long it's been running. What it doesn't: keywords, spend, performance.
Read search ads for offer and proof claims rather than style — a search headline is the most compressed statement of a brand's positioning you'll ever find, because it survived a character limit and an auction.
2. Auction Insights — the free report nobody opens
This is the single most valuable competitor tool in paid search, it's inside your own Google Ads account, and it costs nothing.
Auction Insights tells you, for your own campaigns, ad groups and keywords:
| Column | What it actually tells you |
|---|---|
| Impression share | How much of the available auction each rival is capturing |
| Overlap rate | How often they appear when you do — who you really compete with |
| Position above rate | How often they beat you when you both show |
| Top of page rate | Whether they're buying premium placement |
| Outranking share | The blunt one: who's winning |
Why it beats any spy tool: it's first-party, accurate, and auction-specific. A third-party tool guesses that Brand X competes with you. Auction Insights tells you that Brand X appeared in 64% of your auctions last month and outranked you in half of them — for your keywords, in your geography.
Three things to do with it the first time you run it:
- Segment by keyword, not just campaign. Account-level overlap is an average and hides the keyword where someone is quietly taking you apart.
- Segment by time. A competitor whose impression share climbed over eight weeks has increased budget or improved quality score. That's a change you can act on.
- Compare the list to who you think your competitors are. It is routinely a different list. The brand your founder worries about often isn't in the auction at all, and some aggregator you've never heard of is in 70% of it.
3. Keyword and SERP tools — what they're bidding on
Semrush, Ahrefs, Spyfu and similar estimate competitor keywords, ad copy history and landing pages. Treat the volumes and costs as directional and the copy history as the real asset — seeing how a rival's headlines changed over two years tells you which positioning they abandoned.
Pair this with manual SERP checks in an incognito window, in the geography you actually sell in. Auction results vary by location and device far more than people assume.
4. Shopping and PMax — the feed is public
For ecommerce, competitor pricing, titles and images are visible in Shopping results. Title structure is an underrated steal: the brands winning shopping have usually solved title keyword order, and you can read it directly off the SERP.
5. The seam between the two platforms
Here's a thing worth knowing because almost nobody looks for it: some advertisers run Meta ads whose entire creative is an instruction to go and search a Google keyword.
In the UK car-buying category there are live image ads whose whole creative is a phrase and a prompt — "Search for Sell My Car Instant Valuation UK", live 34 days; "Sell My Car Online", live 56 days; a bare "Sell My Car", live 89 days. No offer, no brand story, just a keyword and an instruction, run by content-arbitrage publishers rather than car buyers.
Those ads are a gift. They name, in public, the exact search terms someone has decided are worth paying to create demand for. If you're in that category, you've just been handed a list of high-value keywords with no tool and no guesswork.
Search your category for this pattern. It's more common in high-intent, high-value categories — finance, insurance, motors, legal — than anyone admits.
A worked example: reading a whole category
Competitor research gets useful when you stop looking at single brands and map the set. "Sell my car" is contested in the UK, the US and India at once, which makes it a good category to read properly.
MotorMoolah takes the hassle angle: selling your car privately? Here's the easier alternative — instant valuation, free collection, direct bank payment, and help settling outstanding finance. Ran 35 days.
CarGurus takes the comparison angle: get multiple offers instantly, and find your best price, with three competing offers shown on the creative. Ran 29 days.
Driveway takes the reassurance angle — selling a car? We've got you covered — and is still running at 5 days in this execution, with earlier siblings at 35 and 11.
Three sensible ads, three mid-length runs. Now set them against what's actually surviving longest in the same category:
| Positioning | Example | Longest verified run |
|---|---|---|
| Narrow qualifier — only 2019+, under 120,000 km | "Why wait weeks to sell your car?" | live 386 days (siblings at 329 and 147) |
| Speed and cash | "Sell your car in 30 minutes, instant cash" | live 155 days |
| Maximum value | "Max cash. Max value." | live 130 days |
| Dealers compete for your car | carwow — "sell it for what it's really worth" | live 109 days |
| Keyword arbitrage | "Sell My Car" as the entire creative | live 89 days |
| Avoid private-sale hassle | MotorMoolah | 35 days |
| Compare offers | CarGurus | 29 days |
Three findings fall straight out of that table, and none required a paid tool.
The longest-running ad in the category is the one that excludes the most people. The 386-day ad only wants cars from 2019 or newer with under 120,000 km on them. It is the narrowest offer in the set and it has outlasted everything else by more than a year. The qualifier isn't a limitation — it's why the ad still works: every click it buys is from someone it actually wants.
The crowded middle runs shortest. Hassle-avoidance and offer-comparison — the two most sensible, most obvious positions — produce the 35- and 29-day runs. That's what entering a saturated angle looks like from outside.
Nobody owns the emotional objection. Every ad in the set is functional: fast, cash, no haggling, best price. The anxiety of being ripped off on the biggest second-hand transaction most people ever make is unclaimed.
Offer wars are visible too
When two brands in a category escalate, the ad library records it like a transcript.
Blue Apron: come back for 30% off your next three orders, code COMEBACK30. Ran 17 days — a reactivation offer aimed at lapsed customers.
HelloFresh, answering in dollars: get $200 off plus free shipping for two months. Ran 6 days.
And HelloFresh again, in a different currency entirely: this is the offer you've been waiting for — 10 free meals plus free breakfast for life. Still running at 13 days, with two near-identical copies.
Two reads. First, both brands are escalating — bigger numbers, shorter runs, which is what an offer war looks like from outside. Second, and more useful: the structure the brand is actively duplicating isn't priced in currency at all. It's priced in meals, a unit the buyer already understands and can picture. The percentage offer ran 17 days, the dollar offer ran 6, and the one counted in meals is the one getting copies built behind it.
If you're in this category, that's your brief: stop discounting in rupees or dollars and start discounting in weeks, meals or boxes.
The three signals that are noise
- Ad count. A brand with 400 live ads isn't winning; it may just be running dynamic creative or a feed. Count distinct concepts, not ads.
- Production value. A competitor's polished film tells you about their budget, not their results. Several of the longest-running ads in this article are text on a background.
- Any "estimated spend" figure. Not available to anyone. Modelled from impressions and a guessed CPM, and wrong at exactly the moments you'd want it.
The weekly ritual — 30 minutes
Monday, what's new. New concepts from your five tracked rivals. Concepts, not ads.
Monday, what survived. Anything that crossed 60 days this week. That's a verdict — log it.
Monday, what changed. Offers, prices, subsidy figures, landing pages. Read the copy, not the picture. The V-Guard example above is invisible to anyone skimming thumbnails.
Monthly, the auction. Auction Insights by keyword and over time. Who's gaining?
Always, end at a brief. A board that never produces an ad is a hobby.
Where the line is
Everything in this playbook uses public ad libraries, your own account's reports, and licensed third-party tools. That's research, and every platform publishes these libraries deliberately.
What's not on the list: creating fake accounts to access gated content, scraping in breach of terms, clicking competitors' ads to cost them money, misrepresenting who you are to get a sales demo, or lifting a competitor's creative rather than learning from it. The first two are contractual problems; the third can be a legal one; the fourth is just bad work that produces a worse ad.
The useful boundary: study the structure, never copy the execution. The 386-day insight is "qualify hard and serve the segment you actually want". That transfers to a clinic, a lender or a SaaS. Their actual ad doesn't.
Where Marxx fits
Marxx tracks competitors across Meta and Google in one place — creative, longevity, copy counts and the keywords they're bidding on — and decodes every ad by hook type, offer type, ad type, emotion, social proof type, authenticity type and intended audience, so "what changed this week" is a filter rather than a memory test.
It also marks each hook, format and emotion in your sector as emerging, rising, plateaued or declining, by how many brands have adopted it and how long their ads stay live — never by spend, which nobody can see.
Book a demo — bring your top five rivals and we'll do a live teardown of everything they're running, on the call. The creative intelligence page explains how each ad gets decoded.
FAQs
How can I see my competitors' Google Ads?
Start with the Google Ads Transparency Center for their live creative, then open Auction Insights in your own account for impression share, overlap rate and outranking share. The first shows what they say; the second shows whether you actually compete.
Is Auction Insights really free?
Yes — it's a standard report in Google Ads, available at campaign, ad group and keyword level. It's first-party and auction-specific, which makes it more accurate than any third-party competitor estimate.
Can I see how much a competitor spends on ads?
No. Nobody can. Figures labelled "estimated spend" are modelled from impressions and an assumed CPM. Use longevity and copy count instead — both are observable.
How do I find which keywords competitors bid on?
Keyword tools estimate it, Auction Insights confirms who shares your auctions, and manual SERP checks in your actual geography verify it. Also watch for social ads whose entire creative is an instruction to search a phrase — those name the keyword outright, and in the car category above they run for months.
Is spying on competitor ads legal?
Using public ad libraries, your own account reports and licensed tools is standard practice. Fake accounts, terms-breaching scraping and click fraud are not. Copying creative is a separate risk from researching it.
How many competitors should I track?
Five. Enough for a pattern, few enough to review weekly. Pick them from Auction Insights rather than from memory — the real list is usually not the assumed one.
How often should I check competitor ads?
Weekly for new concepts and offer changes, monthly for auction share. Daily checking produces anxiety and no additional signal.
What's the single best signal to watch?
How long an ad has been running. It's free, public, and the only performance proxy nobody can fake — the car-buying ad at the top of this article has been live for over a year, which tells you more than any dashboard would.