Paid traffic for AI companion sites: ad networks, funnels and CPA math

Paid traffic for AI companion sites mostly comes from adult ad networks such as TrafficJunky, ExoClick, TrafficStars and JuicyAds, because Google and Meta restrict or block sexually suggestive chat ads. Profitable campaigns pair cheap tier-1 inventory with a bridge page (chat preview, quiz or character picker) and a CPA bid set from your real payer value, not from network averages.

This playbook covers which channels will actually accept your ads, which networks and formats to test first, what the landing flow should look like, and how to do the CPA math before you spend. Every external figure links to its source. Where we use an example, the assumptions are labeled so you can swap in your own numbers.

What paid traffic works for AI companion sites?

Three kinds of paid traffic reach AI companion buyers, and they behave very differently:

• Adult ad networks. Self-serve auctions selling popunders, banners, native, push and video pre-roll on adult tube, cam and content sites. This is where most companion traffic is bought, because the audience is already in an adult mindset and the networks accept adult offers.
• Mainstream platforms (Google, Meta and similar). Huge reach and precise targeting, but strict policies on sexual content and companionship services. Usable only with very clean, non-sexual creative, and even then accounts are fragile.
• Direct and affiliate buys. Flat-rate placements bought from individual site owners, newsletters or communities, plus affiliates you pay on CPA or revenue share. Slower to set up, but often the most stable once you find a placement that converts.

Start with one adult network, prove a funnel converts, then add direct buys and affiliates. Mainstream platforms are a side bet, not a foundation.

Can you advertise an AI girlfriend site on Google or Meta?

Only in a narrow, sanitized form, and it is risky to build a business on it.

Google introduced a standalone Dating and Companionship Ads policy with enforcement starting March 4, 2025. Advertisers in scope must be certified by Google before running ads. The policy prohibits promoting paid companionship or sexual acts and imagery that does not represent the users a service markets to. It restricts livestream or chat apps featuring nudity, partial nudity or sexually suggestive content, and racy imagery in ads or on landing pages. Restricted ads can be limited by user age, local law, SafeSearch settings and the user's search queries, which in practice means small reach for most companion offers.

Meta's ad standards on adult nudity and sexual activity are stricter. Ads must not show nudity or near nudity, sexual activity, sexually suggestive positions, or logos and clips from known pornographic sites. Imagery of people in sexually suggestive poses or revealing clothing has to be limited to users over 18, even when it is AI-generated.

Enforcement has been uneven, which tempts some advertisers to push the line. In 2024 Wired found more than 29,000 ads for AI girlfriend apps in Meta's ad library, many of them labeled NSFW, and Meta said it was removing the violating ones, as Fox Business reported. That is exactly the risk: ads that slip through today get accounts, pages and payment methods banned tomorrow. If you test mainstream platforms, run clean creative that sells the companionship and conversation, not the body, and send it to a non-explicit landing page.

Which adult ad networks should you test first?

The four networks most companion advertisers start with are TrafficJunky, ExoClick, TrafficStars and JuicyAds. TrafficJunky sells inventory on Pornhub and its network and says Pornhub alone has over 133 million daily visitors, with a self-serve dashboard and CPM bidding. The others sell across thousands of independent adult sites.

Networks rarely publish their minimums and price floors, so the figures below come from a third-party roundup by AFFNinja, an affiliate publisher. Treat them as starting ranges and confirm in each network's dashboard before you fund an account.

The same roundup puts tier-2 and tier-3 countries well below these numbers, often under $2 and sometimes under $1 per thousand impressions. Cheap countries are useful for testing whether a pre-lander gets clicks, but payer value usually falls faster than CPM does, so judge every country by cost per payer, not cost per click.

How to choose your first source: a comparison of TrafficStars, TrafficJunky and ExoClick recommends TrafficStars as a first test for AI dating funnels because of its format flexibility, ExoClick for broad format testing and lower-cost countries, and TrafficJunky once your creative is polished and you are buying tier-1 traffic. In short: learn cheaply, then pay for premium inventory once you know what converts.

Which ad formats convert for AI companion offers?

Each format brings a different level of intent, so the right one depends on what your landing flow asks the visitor to do.

• Popunder. Cheapest volume and the fastest way to get data. Visitors did not click anything, so intent is low. Works only with a strong bridge page that earns the click.
• Native. Looks like content, costs more per click than pops, and pre-qualifies users because they chose to click a headline and image. Good for story or curiosity angles.
• Display and banner. Reliable on premium placements and good for retargeting-style messages to people who already know the brand or the character.
• Push and in-page push. Short, chat-style notifications. They can perform well for companion offers but fatigue quickly, so plan to rotate creative every few days.
• Video pre-roll. The most expensive CPMs in the table above, but a short clip of a character and a chat screen explains the product faster than any banner.

A sensible first test is one popunder campaign for fast data and one native or banner campaign for higher intent, sent to the same two pre-landers, so you learn about both the format and the page at once.

What should the landing page for paid traffic look like?

Do not send cold paid traffic straight to a signup wall or a gallery. The AFFNinja comparison advises a warm-up step such as a quiz, an avatar selector, a comparison page or a chat preview before the paywall. The logic is simple: the product is conversation, so let visitors feel the conversation start.

A chat-first paid funnel usually has four steps:

1. Age gate. A clear 18+ confirmation, plus any verification your jurisdiction requires, before anything adult appears.
2. Bridge page. A character picker, a two or three question quiz ("What kind of conversation are you in the mood for?") or a short preview of a chat thread. The goal is a small commitment before you ask for anything.
3. First messages free. Visitors chat with the character they picked and see that it remembers what they said. This is the moment that sells the product.
4. Paywall at the natural break. Ask for a subscription or a token pack when the visitor wants more: a longer conversation, a personalized image or a locked scene.

Keep the promise consistent from ad to page. If the ad shows a specific character, the bridge page should show that same character. Google's policy explicitly prohibits imagery that does not represent the users or experience a service offers, and adult networks reject misleading angles too. The AFFNinja comparison specifically warns against fake chat requests, celebrity imagery, deepfake or undressing angles and misleading claims. On top of the policy risk, a mismatch between ad and page drives chargebacks, which threaten your payment processing long after the campaign ends.

Creative rules that keep accounts alive

• Adults only, visibly. Every character in every ad should read clearly as an adult. Avoid school settings, youthful styling and any wording that implies otherwise.
• No real people. No celebrities, creators or lookalikes unless you hold a written license, and never sexual content of a real person without it.
• Sell the experience. Hooks that show the chat ("She remembers your bad day") travel further across networks than explicit imagery, and they survive policy reviews.
• Test openers, not just images. Run two versions of the first message a character sends, for example playful versus attentive, and measure which one gets the visitor to reply.
• Rotate before fatigue. Push and popunder audiences see the same creative many times. Refresh hooks on a schedule rather than waiting for click-through to collapse.

How do you calculate CPA and a break-even bid?

Paid traffic is a math problem. You need three numbers before you spend: what a paying member is worth to you, what share of that you keep, and how many visitors it takes to make one payer.

Start with conversion. RevenueCat's State of Subscription Apps 2026 found that apps asking for payment upfront convert a median 10.7% of downloads to paid by day 35, against 2.1% for freemium, and that AI-powered apps earn 41% more revenue per payer but churn 30% faster. Those are app benchmarks, not web companion sites, but they show the trade-off: an earlier ask converts more, and retention is what decides whether a payer is worth the cost.

Then track every step of the funnel, not just the final CPA. The ladder we recommend:

A worked example (assumptions labeled)

Assume a $1,000 test on US popunder traffic at a $3 CPM, inside the reported range above. That buys about 333,000 impressions, and with a popunder each impression is a page view. Assume 3% of viewers engage with the bridge page (10,000 people), 10% of those start a chat and create an account (1,000), and 4% of signups buy (40 payers). Your CPA is $1,000 / 40 = $25.

Now the value side. Our AI companion site earnings calculator uses $30.23 in average monthly revenue per paying member as a planning input. If a payer stays three months on average, they generate about $90.69 in gross site revenue. At a revenue share of up to 60%, your cut is up to $54.41, which leaves up to $29.41 per payer after the $25 CPA. That is a working campaign. If the same payers stayed only one month, your cut would be up to $18.14 and every payer would lose money.

The break-even rule is simple: maximum CPA = average gross revenue per payer x your revenue share. Set your target CPA well below that line during testing, because early cohorts always look better before refunds and churn arrive. Our ARPU and LTV benchmarks explain how to build that payer value from your own cohorts instead of averages.

How much budget do you need to test paid traffic?

Enough to get a statistically useful number of payers from each variant, not just clicks. The AFFNinja comparison suggests a $500 to $1,500 test budget and a starting structure of one traffic source, one offer path, two pre-landers and three to five countries. Fewer variables make results readable.

Practical rules for the first test:

• Decide kill rules in advance. For example, pause a placement once it spends twice your target CPA without a payer.
• Split mobile and desktop. They price and convert differently, and blending them hides your best segment.
• Use server-to-server postbacks. Send signups and purchases back to the network with sub-IDs per placement, so the network's optimization works on payers, not clicks.
• Build whitelists and blacklists. After a few days, keep only the placements that produce engaged visitors and block the rest. Placement pruning is the main lever you control on a self-serve network.
• Give each variant time. Judge on payers and refunds over at least a week, not on day one click-through.

How do you scale a profitable campaign?

Once one source, one country and one funnel hit your target CPA for a couple of weeks, scale in this order: raise bids on whitelisted placements, add the next country with similar payer value, add a second format on the same network, then add a second network. Each step changes one variable, so you know what broke if CPA jumps.

Retention is the other half of scaling. Every extra month a payer stays raises your maximum CPA, which lets you outbid competitors for the same inventory. Welcome sequences, new characters and content drops do more for paid traffic economics than any bid tweak. Our AI fan site retention playbook covers the levers.

This is also where the operator model matters. On a WhiteLabelFans site, approved partners focus on traffic and keep up to 60% of total site revenue, including subscriptions, tips, content unlocks and upsells, while the platform runs hosting, billing and compliance. That lets a media buyer put their time into the part of this playbook that moves margin: creative, placements and funnels.

Frequently asked questions

What is the best ad network for AI companion sites?

There is no single best network. TrafficStars and ExoClick are common first tests because they offer many formats and lower-cost countries, while TrafficJunky sells premium Pornhub inventory that suits polished creative and tier-1 traffic. Start with one network, prove a funnel converts, then add others. Judge every source by cost per payer, never by CPM or click-through alone.

Can I run AI girlfriend ads on Facebook or Instagram?

Only with non-sexual creative. Meta's ad standards ban nudity, near nudity and sexually suggestive positions, and require 18+ targeting for revealing imagery, including AI-generated images. Many AI girlfriend ads have run anyway, but Meta removes violating ads and bans repeat offenders. Treat Meta as a side test with clean, conversation-focused creative, not as your main traffic source.

How much does adult traffic cost?

Third-party roundups put US CPMs on major adult networks roughly between $1.50 and $12 depending on format, with popunder and push at the low end and video pre-roll at the high end. Tier-2 and tier-3 countries are often under $2. These are estimates; check live bid ranges in each network's dashboard before you plan a budget.

What is a good CPA for an AI companion site?

A good CPA is one below your break-even line: average gross revenue per payer multiplied by your revenue share. If payers generate $90 over their lifetime and you keep a share of up to 60%, your ceiling is up to $54. Aim well under that while testing, because refunds and early churn lower real payer value.

Should paid traffic go straight to the signup page?

Usually not. Cold adult traffic converts better through a bridge step such as a character picker, a short quiz or a chat preview, followed by a few free messages before the paywall. The bridge earns a small commitment, filters out low-intent clicks and reduces refunds. Keep the character and promise in the ad identical to what the page shows.

Sources

• Google Ads: New Dating and Companionship Ads policy and advertiser certification program
• Meta Transparency Center: Ad standards, adult nudity and sexual activity
• Fox Business: Ads on Facebook, Instagram for explicit AI girlfriends prompt Meta crackdown
• TrafficJunky: CPM advertising network
• AFFNinja: Top adult ad networks and platforms
• AFFNinja: TrafficStars vs TrafficJunky vs ExoClick compared
• RevenueCat: State of Subscription Apps 2026