What's happening with Facebook ads in 2026: why lead quality dropped and what to do about it

“We're getting more leads but fewer customers.” In 2026 we hear this from almost every new client. Someone left a phone number and doesn't remember doing it. The number doesn't answer. People ask about things you don't sell. Meanwhile, the cost per lead in the ad account looks fine, and it seems the ads are working.

Let's look at what has actually changed in Meta advertising, why lead quality is falling and how to bring it back, using the steps we apply in our projects.

What changed: AI now runs the ads

Just a couple of years ago a paid social specialist picked interests, age and location by hand, and those settings decided who saw the ads. Now Meta's algorithms do most of the work:

• Andromeda is the ad retrieval system. At the moment of delivery it picks a few thousand ads relevant to a specific person out of tens of millions.
• GEM is a large model that predicts who will take the desired action. According to Meta, after it was rolled out in Q2 2025, conversions rose 5% on Instagram and 3% in Facebook Feed.
• Advantage+ means automated campaigns where your audience settings act more as a suggestion than a constraint.

At the same time, competition for impressions is growing. According to Meta's Q2 2026 report, the average price per ad rose 12% year over year and the number of impressions rose 14%. Ads are getting more expensive, and the algorithm tries even harder to hit your goal as cheaply as possible.

The key takeaway: the algorithm doesn't know who your customer is. It only knows the event you train it on. Train it on “filled in a form” and you'll get people who like filling in forms. Train it on “became a customer” and you'll get customers.

6 reasons leads got worse

1. Optimizing for leads, not customers

This is reason number one. The “Leads” objective tells the algorithm: find people who will fill in the form most cheaply. It does. Some of them are your customers, but many click on everything, enter giveaways and leave their details “just in case”.

2. Pre-filled forms

By default, Meta's lead form fills in the name and phone number from the profile, and people submit in two taps, sometimes without reading what they're signing up for. Hence the “I never left anything” replies and old numbers that are no longer in use.

3. Audience Network placements

With automatic placements, some impressions go to third-party apps and games. There are many accidental taps and reward-driven clicks there. For lead generation it's one of the “dirtiest” sources.

4. A creative everyone likes

Targeting settings used to filter out the wrong audience. Now audiences are broad, and the creative itself is the filter. If the ad doesn't say who the product is for, what it costs and what happens after you sign up, everyone who's curious will fill in the form.

5. Leads are handled too late

A social media lead cools off fast: people leave a request in passing and an hour later don't remember where. Back in 2011, Harvard Business Review showed that companies that contacted a lead within an hour were nearly 7 times more likely to have a qualifying conversation than those that called back an hour or more later. Today, in the age of messengers, the window is even shorter.

6. Bots and fake leads

Fake leads are a real problem, especially in niches with a high cost per lead. You can spot them by identical names, numbers from other countries and batches submitted at the same time. The fix is in form and placement settings, covered below.

Diagnose first: is it the ads or the sales?

Before changing your campaigns, sort the last 100 leads by status. It takes an hour but saves weeks of blind experiments.

Here's what a healthy funnel looks like. In the case of the kids' speech therapy school Mova Promova, 5,358 of 8,930 leads made it to a paid assessment, or 60%, and 44% became the school's customers. These are the numbers we watch, not the cost per lead in the ad account.

8 steps that bring lead quality back

1. Switch on the “Higher intent” form. It adds a review screen before submission. You'll get fewer leads, but they'll be people who deliberately pressed “submit”.
2. Add 1–2 qualifying questions. The child's age, city, budget, timing: anything that filters out people you definitely aren't right for. Don't turn the form into a survey: two questions work, eight kill conversion.
3. Send CRM statuses to Meta. Connect the Conversions API and report what happened to each lead: “qualified”, “booked”, “bought”. This is the key step: the algorithm starts learning from your customers rather than from form-fillers.
4. Switch the objective to conversion leads. Once CRM data is flowing, choose optimization for quality leads. According to Meta, such campaigns deliver on average about a 15% lower cost per quality lead and a 44% higher share of leads that become quality leads.
5. Turn off Audience Network for lead generation, or at least compare lead quality from that placement separately.
6. Make the creative a filter. Name the audience (“for parents of kids aged 4–7”), the price or price range and what happens after sign-up. The more specific the message, the fewer random people.
7. Handle leads within minutes. An automatic messenger reply right after the request, then a call or message from a manager within 5–15 minutes. Every hour of waiting lowers your contact rate.
8. Track cost per qualified lead and per sale. Cost per lead in the ad account is an intermediate metric. A campaign with $8 leads that are 50% qualified beats one with $4 leads that are 15% qualified.

Lead form or website?

A common question. The short answer: test both. A lead form brings more leads, usually more cheaply; a website brings fewer, but people have time to read about the product and arrive better informed. In high-ticket niches with long decisions (education, healthcare, B2B), a website or quiz often wins on cost per sale even if it loses on cost per lead. In niches with quick decisions, the form is usually more effective. Compare only on cost per qualified lead and per sale.

What not to do

• Don't narrow the audience to specific interests hoping to filter out the wrong people. In 2026 this more often raises the price than improves quality: the algorithm needs room, and it's better to filter with the creative and the form.
• Don't change everything at once. Change the form, creatives, objective and budget on the same day, and you won't know what worked.
• Don't judge by the first few days. After you change the optimization goal the algorithm relearns, and metrics may fluctuate for the first 2–3 weeks.

Sources

1. Meta: Second Quarter 2026 results
2. Meta Engineering: Andromeda, the next-gen ad retrieval engine for Advantage+ (2024)
3. Meta Engineering: the GEM model, +5% conversions on Instagram and +3% in Facebook Feed (2025)
4. Click Here Digital: Meta's shift to conversion leads optimization (Meta data: −15% cost per quality lead)
5. Harvard Business Review — The Short Life of Online Sales Leads (2011)