Build Your First AI Project This Weekend
Stop consuming tutorials. Start creating. Get the free step-by-step guide.
2026 Update:
Popunders are still alive. Browsers have been trying to kill them for 20 years. The Coalition for Better Ads banned desktop and mobile pop-ups in 2019. In 2026, PropellerAds still claims more than 750 million popunder impressions a day, and Adsterra claims 12 billion pop impressions a month (AffMaven).
AI helps you write landing pages and cut bad placements faster.
But this is still a media-buying channel. It is not how a dentist, plumber, or lawyer should buy customers.
If you run a local service business, start with Google Ads and local SEO. People searching “emergency plumber near me” have intent. People who just got a popunder do not.
This guide is for affiliates, CPA buyers, and operators who already understand tracking.
Who pop traffic is for (and who should skip it)
Pop traffic is interruptive. Someone is on a torrent site, a file host, a scraped blog, or a parked domain. Your page opens behind or instead of what they wanted.
Use it if:
- You buy CPA / CPI / SOI offers and can blacklist placements
- You have a tracker and can read source IDs
- You can lose a few hundred dollars learning without blinking
- Your offer is one click, one form, one install
Skip it if:
- You are a local service business
- You care about brand reputation
- You need high-intent buyers
- You do not have conversion tracking
I have bought pops since 2015. Cheap volume is real. Clean leads for a dental office are not.
What pop traffic actually is
Three formats:
Popunders. A new window or tab opens behind the current page. The user sees it later. This is most of what you buy.
Pop-ups. Opens on top. Browsers block most of this. Avoid.
Redirect / domain traffic. Parked or typo domains. Someone types a dead URL and lands on your offer. Slightly more intent than a random popunder, still not search intent.
The visitor did not ask for you. You have a few seconds. Simple offer, fast page, obvious hook. No 2,000-word sales letter.
Offers that still convert (and what died)
Impulse, low friction. If the offer needs a sales call, use search.
Sweepstakes / SOI. Email or phone for a chance to win. Easy to launch. Payouts are modest. Compliance is tighter than 2018. Run real offers from a real network. Fake “you won” pages get you banned and can get you sued.
VPN and utilities. Privacy, cleaners, extensions. Evergreen demand. Angles are saturated. You win on page speed and placement lists, not on a clever headline.
App installs (CPI). Games, tools, social apps. Demand is consistent. Quality caps are real. Networks will cut you if installs bounce.
What I would not start with in 2026:
- Fake antivirus / “your phone is infected”
- Pin-submit / carrier billing in Tier 1 geos (carriers killed most of it)
- Anything that requires a credit card on the first page
- Local services (“dentist near you” on a popunder is a waste of money)
Start with a sweepstakes or VPN in a cheaper geo if you are learning. Not because it prints money. Because the feedback loop is fast.
Networks that still exist in 2026
I am not going to dump a graveyard of 2016 brands. If a network is not clearly operating, it is not here.
Learning budget (under $500)
- $10 minimum deposit, $2.50/day campaign floor (AffMaven)
- Popunder only. No push, no native.
- Blind network: you see placement IDs, not URLs
- Best as a classroom, not as a career
Full walkthrough: PopAds guide.
Useful tokens: [WEBSITEID], [QUALITY], [ADBLOCK], [BROWSERNAME], [OSNAME], [ISPNAME], [SCREENRESOLUTION]
Volume after you have a converting offer
- Big pop + push inventory
- CPA Goal bidding if you can pass conversions back
- Better anti-fraud than budget pops
- Slower approval, higher prices
- Direct publishers plus exchange traffic
- Pop, push, native, display
- You still blacklist. “Direct” is not “clean.”
- AdScore filtering
- Traffic tiers (premium vs remnant)
- You pay more. You waste less. Fair trade if you have already burned money on junk.
Other live options
- HilltopAds - gambling, dating, finance, adult
- ExoClick - adult verticals
- PopCash - cheap, heavy file-sharing. Fine for tests, not for quality
- Adcash - multi-format, easy to overspend if you do not cap variables
- Clickadu - stricter review, SmartCPM
Skip names you remember from 2015 forum threads unless you can log in today.
Redirect (often overlooked)
ZeroPark - domain, push, pop. Same company family as Voluum, so tracking is less painful.
Tokens: {keyword}, {target}, {match}, {cid}
Redirect is still “they typed something.” That is more intent than a popunder from a torrent site. Still not Google Search.
How to launch without lighting the budget on fire
1. Get an offer from a real network
Ask a manager at a CPA network what is converting on pop right now. Ask geo, payout, creative rules, and whether they will bump payout for quality. Get two offers in the same vertical.
If you do not have a network yet, read affiliate marketing with AI first. Do not buy traffic for an offer you cannot track.
2. Track everything
No tracker, no campaign.
Pass every token the source gives you. Details: tracking guide.
3. Spy, then steal the structure (not the scam)
Adplexity or Anstrex. Look at layout, language, CTA, and funnel length. Do not copy fake urgency and fake virus pages.
Feed 10 live pages to Claude:
I am launching a [OFFER TYPE] pop campaign in [GEO].
Here are the top landing pages I found (describe layout, headline, CTA, form fields).
Tell me:
1. Shared layout patterns
2. Hooks that are still compliant
3. What to drop because it will fail review
4. A simple page structure I can build today
4. Build a stupid-fast page
One headline. One image. One button. Mobile first. Under two seconds. Guide: landing page code.
5. Launch narrow
- $50-100/day
- One geo
- Mobile, Android, Chrome to start
- Frequency cap 1-3 per user per 24 hours
- Do not touch it for 24-48 hours
6. First cut at 48 hours
Blacklist placements with spend and zero conversions. Raise bids on placements with conversions near payout. Kill the campaign if you have spent 2-3x payout with no signal.
Signs to scale: conversions near target CPA, repeat placements converting.
Signs to kill: random noise, no conversions after a few hundred dollars, CTR with zero on-page time.
Tracking tokens (save these)
If the token is not in the click URL, you cannot blacklist later. That is the whole optimization model.
PopAds: [WEBSITEID], [QUALITY], [ADBLOCK], [BROWSERNAME], [OSNAME], [ISPNAME], [SCREENRESOLUTION]
PopCash: %siteid%, %cc%
ZeroPark: {keyword}, {target}, {match}, {cid}
Map them to the same columns in your tracker (source, keyword, country) so a Monday export looks the same no matter which network printed the click.
Frequency and hours
Cap at 1-3 per user per day unless you have data that a second view converts. Pop users did not opt in. The third impression is usually anger.
Daypart after you have conversions by hour, not before. A “7pm-7am” rule is a starting guess for impulse offers, not a law. Your geo’s lunch break might be the only time real people are on auto blogs.
Mobile page vs desktop page
Do not send both to one URL “because it is responsive.” Mobile popunders are delayed. Desktop tab-ups are immediate. Write two pages. The mobile page should make sense three hours later. The desktop page can be louder.
AI that actually helps (and what it cannot do)
Pop campaigns vomit data: site IDs, OS, browser, ISP, hour, geo. AI is good at ranking that list. It is bad at inventing a business model.
Weekly prompt I actually use:
Campaign data, last 7 days:
[paste placement, spend, clicks, conversions, revenue]
Target CPA: $[X]
Current CPA: $[Y]
1. Blacklist (spend over $[threshold], zero conversions)
2. Raise bids (profit, room to scale)
3. Device / OS / browser cuts
4. Hours to pause
Action list I can implement in 15 minutes.
Fraud check when CTR is high and conversions are zero:
Flag bot / farm / incent patterns from this click log:
[time of click, time on page, conversion time, repeating fingerprints]
List source IDs to kill.
AI will not make torrent traffic buy Invisalign.
Mistakes that still kill campaigns
Starting too broad. All devices, all browsers, all countries. You learn nothing. Start narrow, expand with data.
No frequency cap. Showing the same user 40 pops is how you buy hatred, not conversions.
Optimizing for CTR. Clicks are not money.
Blaming “bad traffic” while running one landing page for three weeks. Always have two pages. Kill the loser.
Using pop to generate leads for a local business. A dentist does not need 10,000 low-intent emails. They need people searching for a dentist. That is Google Ads and the Map Pack.
Ignoring time zones. Your dashboard is in Eastern. The traffic is in Jakarta. Conversions look “overnight” because you misread the clock. Set the tracker to the geo’s local time before you daypart.
Paying for pops on brand terms. If you sell a named product, search captures the people who already want it. Pop captures people who wanted a PDF. Different jobs.
What to expect
Margins are thin. If you are up 20-30% after tracker fees, you are doing fine. The “six figure months” you hear about are volume across geos, not fat margins on one campaign.
Pop is a classroom. Tracking, blacklists, fast pages, ruthless cuts. Those skills transfer. The traffic quality does not transfer to a service business.
Next step
No affiliate network yet: affiliate marketing guide.
Have offers, no tracking: BeMob or the tracking guide.
Ready to spend $50: PopAds walkthrough.
Local service business that landed here by accident: close this tab. Open Google Ads and local SEO.
Related: push traffic | native ads | paid traffic map | RTB
Have questions? Contact me.