Performance Max, Advantage+ and Smart+ in the UAE: keeping AI-run campaigns from buying demand you already own

Digital marketing

All three major ad platforms have converged on the same bet: give the machine one campaign type and let it buy. Run carelessly in the UAE, that machine quietly buys your brand searches, your existing customers and your WhatsApp enquiries, reports a flattering CPA, and leaves real prospecting starved. Here is how to see it, price it and fence it.

Google, Meta and TikTok have all made the same structural move. Google Ads pushes Performance Max, a single goal-based campaign that reaches Search, YouTube, Display, Discover, Gmail and Maps at once (Google). Meta has renamed Advantage+ Shopping to Advantage+ Sales and now switches Advantage+ on by default for new Sales, App and Leads campaigns (Jon Loomer). TikTok's Smart+ automates targeting, optimisation and creative, and its upgraded flow is now the default path for Traffic, Sales, Lead Generation and App Promotion (TikTok Ads Manager).

The pitch is identical on all three: fewer settings, more conversions. The catch is that in this market the failure mode is unusually profitable to ignore. A typical UAE or GCC advertiser holds three assets that make automated campaigns look brilliant while doing damage: a strong bilingual brand with heavy branded search volume, a WhatsApp-first enquiry flow that counts raw conversations as conversions, and a CRM that never talks back to the ad platform. Feed that account to Performance Max or Advantage+ and the machine finds the cheapest conversions available, which are almost always brand queries and existing customers. Reported CPA falls. Blended cost per qualified lead rises. Nobody notices until the pipeline thins out.

One caveat before anything else: all three systems amplify whatever conversion signal you feed them. If your conversions are raw form fills and unqualified WhatsApp threads, you are not automating demand generation, you are automating noise at scale (LGG Media).

The UAE failure mode: paying for demand you already own

Three things happen when an AI-run campaign lands in an unguarded UAE account.

Brand capture. Branded search converts well: practitioner audits put branded query conversion rates at 12 to 20 per cent (Hustle Marketers). Performance Max bids on your own brand terms, often at 2 to 4 times the CPC your dedicated brand Search campaign pays for the same click, and takes the conversion credit. Google does prioritise exact-match Search campaigns over PMax for overlapping queries, but PMax picks those queries up the moment your Search campaign becomes budget-constrained (LGG Media). The leak therefore appears exactly when you scale PMax spend. The same audit work puts it at 10 to 25 per cent of campaign budget where no brand exclusions are set.

Existing-customer recycling. In a market where the same CRM list gets re-marketed constantly, automated campaigns drift toward people who were going to buy anyway: recent purchasers, current enquiry lists, your own staff. Meta removed the existing-customer budget cap for new Advantage+ campaigns, so the single setting that used to cap this is gone; the workaround is an ad-set structure (Jon Loomer, DTC Newsletter).

WhatsApp lead inflation. Click-to-WhatsApp and instant-form campaigns optimising on "conversation started" or "form submitted" buy exactly what those events measure: contact, not intent. Every idle window-shopper satisfies the goal. The campaign reports cheap conversions; your sales team reports tyre-kickers.

The result looks like this: PMax or Advantage+ CPA drops by roughly a third in the dashboard, the sales-qualified lead rate halves, and your brand campaign loses impression share. The machine did not get better at finding customers. It got better at finding your customers.

What each platform controls, and what you keep

Notes on that table. Google is explicit that PMax is built to complement keyword-based Search, not replace it (Google). Meta's Advantage+ Audience controls (location, minimum age, excluded custom audiences, languages) are constraints the system must respect; its audience suggestions are hints it may ignore (Jon Loomer). TikTok's upgraded Smart+ flow lets you mix full automation, partial automation and manual settings module by module, including manual placement selection, and allows up to 50 creative assets per ad (TikTok Ads Manager). All three interfaces change monthly, so verify every setting in your own account at writing time before relying on it.

Worked example: AED 60,000 a month, measured honestly

The figures below are illustrative, but the shape is one we see repeatedly. A Dubai home-services firm (fit-out and maintenance) spends AED 60,000 a month: Google AED 35,000, Meta AED 25,000. Google splits into brand Search (AED 4,000), non-brand Search (AED 22,000) and Display plus YouTube (AED 9,000). Non-brand Search delivers qualified leads at about AED 750 each; the brand campaign exists to defend branded queries at an average CPC around AED 8.

Step 1: the migration. The team consolidates non-brand Search plus Display and YouTube, AED 31,000 a month, into one Performance Max campaign optimising on "lead submitted", a WhatsApp-sourced form fill with no qualification. Six weeks later, PMax reports 310 conversions at AED 100 each. It looks like the best campaign in the account: reported cost per lead has fallen more than sevenfold.

Step 2: the leak check. Three diagnostics, in order:

• The brand Search campaign's impression share drops from 78 per cent to 52 per cent. Nothing else in that campaign changed.
• Search terms insights for the PMax campaign show brand-named categories carrying roughly 38 per cent of conversions (Google).
• Sales reports that one PMax enquiry in eight becomes a qualified lead, against one in three from the old non-brand Search enquiries.

Step 3: the honest number. Cost per qualified lead, not cost per raw lead. Before the migration, AED 31,000 bought roughly 41 qualified leads a month at about AED 750 each. Under unguarded PMax the same spend produces roughly 39 qualified leads, about AED 800 each: the reported sevenfold CPA improvement is an illusion, because the raw leads did not qualify. Strip out the conversions that carried brand categories or matched existing contacts, roughly 40 per cent of the total, and the genuinely incremental cost per qualified lead is closer to AED 1,350.

Step 4: the guardrails. Apply the checklist below: own-brand and competitor brand exclusions on every PMax campaign, an account-level negative keyword list (careers, login, support, free and the usual), location set to Presence rather than "presence or interest", Customer Match exclusion of the client base and the enquiry list, and weekly offline conversion imports of qualified-lead events with values attached to the click IDs.

Step 5: the re-measure. Reported PMax CPA rises from AED 100 toward AED 200: the machine lost access to the cheap conversions it was claiming. Brand impression share recovers within weeks. With the conversion signal fixed, qualification climbs back toward one in three, so the same budget returns qualified leads at or below the AED 750 the account managed before the migration. Practitioner audits report exactly this pattern: reported campaign ROAS falls, blended account performance improves over two to three weeks (Hustle Marketers).

The lesson: the first number that moves when you add guardrails is the reported one, and it moves the wrong way. Judge on blended cost per qualified lead, never on campaign-reported CPA.

The guardrail checklist: levers that actually bind

Google (Performance Max)

• Brand exclusions, campaign level. Build a brand list (your brand plus sub-brands and common transliterations) in the shared library, then apply it to every PMax campaign. Google matches brand entities including known misspellings and variants, which is stronger than text-based negatives; you can also keep Shopping ads serving on excluded brand terms with one tickbox (Google). A second list holds direct competitors you do not want to subsidise.
• Account-level negative keywords. These apply to Search and Shopping inventory across all campaign types, including PMax (Google). Cover careers, jobs, login, support, salary and anything else that flatters your raw-lead counts.
• Location on Presence. Set location targeting to Presence, not the default "presence or interest", so you are not paying for people merely reading about Dubai (LGG Media).
• Audience exclusions. Exclude Customer Match lists of current clients, staff and active enquiries from prospecting campaigns.
• Bid on qualified leads, not raw fills. Import offline conversion events (SQLs, qualified calls, closed deals with values) tied to GCLID or WBRAID. Practitioner guidance is blunt: below roughly 15 to 30 qualified conversions per campaign per month, Smart Bidding is guessing across six channels, and a daily budget of roughly 3 to 5 times target CPA is the floor for the learning phase to function (LGG Media).
• Watch, do not trust. Pull search terms insights monthly. A GAQL query for the API, run per campaign:

SELECT
  campaign_search_term_insight.category_label,
  campaign_search_term_insight.id,
  metrics.clicks,
  metrics.impressions,
  metrics.conversions,
  metrics.conversions_value
FROM campaign_search_term_insight
WHERE campaign_search_term_insight.campaign_id = 1234567890
  AND segments.date BETWEEN '2026-09-01' AND '2026-09-30'
ORDER BY metrics.conversions DESC

Brand-named categories should show minimal volume. If they persist, the exclusion is not catching the entity and you need the negative keyword backup.

Meta (Advantage+)

• Use audience controls as constraints, not suggestions. Advantage+ Audience lets you set location, minimum age, languages and excluded custom audiences as hard controls; age maximum, gender, custom audiences and detailed targeting are suggestions (Jon Loomer). Exclude your purchaser and current-enquiry custom audiences everywhere you prospect.
• Rebuild the existing-customer cap. With the budget cap gone from new Advantage+ campaigns, the accepted pattern is two ad sets sharing one Advantage+ campaign budget (Meta): one ad set including existing customers, with a spend cap, and one broad ad set excluding them (Jon Loomer).
• Check the audience breakout weekly. The Breakdown dropdown in Ads Manager splits spend and results by new, existing and engaged customers (Foxwell Digital). If the existing-customer share of spend is climbing, your acquisition campaign is a retention campaign with better reporting.
• Fix the WhatsApp signal. If your conversion is "conversation started", qualify it: import the qualified-lead event through the Conversions API with values, so the campaign bids on the enquiry that reached a real conversation, not the one that sent "hi" and vanished. A minimal definition worth writing down before any tooling:

{
  "event_name": "qualified_lead",
  "source": "whatsapp_enquiry",
  "qualifiers": ["service_area_confirmed", "budget_stated", "timeline_within_90_days"],
  "value_aed": 1200,
  "identity_fields": ["click_id", "hashed_phone"],
  "import_target_hours": 72
}

TikTok (Smart+)

• Use the upgraded flow's controls. Automatic placement is the default, but the upgraded experience allows Select placement, audience controls the system enforces, or manual targeting (TikTok Ads Manager). In the UAE context, enforce language and location controls; treat the rest as flexible.
• Exclude customer audiences from prospecting ad groups using customer file or engagement audiences, as with Meta.
• Read creative-level reporting. The upgraded flow reports performance grouped by creative material ID, which is where you find your Arabic-versus-English winner (TikTok Ads Manager).
• Qualify the event before scaling. Same rule as everywhere: optimise on Events API-backed qualified-lead events with values, never raw clicks or form opens.

Creative supply is the real bottleneck

Here is the part that surprises finance directors. The constraint on Advantage+ and Smart+ performance is rarely bidding logic; it is creative volume and variety. TikTok's Smart+ wants combinations of video, text and CTAs, with up to 50 creative assets per ad in the upgraded flow (TikTok Ads Manager). Meta's automated delivery burns through creative combinations quickly and quietly reallocates spend to whatever retains attention. Performance Max asset groups perform best with varied headlines, descriptions, images and video, though retail accounts with a Merchant Center feed can launch without custom assets (Google).

In the UAE, bilingual doubles the problem. Every concept needs an Arabic and an English execution, and short-form vertical video for TikTok and Reels is a separate production line from static display. TikTok's Symphony automatic enhancements cover resizing, quality boosts, translation and dubbing, which helps with volume (TikTok Ads Manager), but the concepts still have to exist. Plan creative as a monthly production quota per campaign, with a 9:16 video pipeline as the backbone, or the automation will optimise a shrinking pool until it exhausts it.

Seasonality: the machine optimises on the recent past

Automated bidding learns from trailing data, and in the UAE that matters twice a year.

Ramadan and Eid. Ramadan moves about 10 days earlier each year: the 2026 cycle started in mid-February, and the 2027 cycle is expected to begin in early February, moon sighting permitting. Enquiry volume shifts to night hours, post-Iftar decision windows compress in the days before Eid al-Fitr, gifting and hospitality categories spike while B2B pipelines slow. An automated campaign left alone will keep bidding as if February is January. Move budgets, dayparts and creative themes manually, ahead of the shift, and exclude seasonal terms you do not sell into.

Summer travel. From late June through August a large share of resident demand leaves the country, while travel, hospitality and airport-adjacent categories surge. Same principle: the Smart Bidding layer chases the trailing 30 to 90 days of conversions. Set the seasonal plan explicitly, move budget in advance, and expect your "improvements" to look worse for two weeks after each shift.

The guardrails do not change with the season. Brand exclusions, customer exclusions and clean conversion signals matter more during demand shifts, not less, because that is when the automated system is most tempted to harvest whatever demand still exists: yours.

Do this in the next 30 days

Week 1: measure the leak. Export 90 days of brand Search impression share. Pull search terms insights for every PMax campaign (Google). In Meta, open the audience breakout on every Advantage+ campaign and record the existing-customer share of spend (Foxwell Digital). Compute blended cost per qualified lead, not cost per raw lead. Change nothing yet.

Week 2: apply the hard guardrails. Brand lists (own brand, then competitors), account-level negative keyword lists, Presence-only location, Customer Match or custom audience exclusions of clients, staff and active enquiries. Google's brand exclusions take effect within a day or two; expect reported campaign performance to dip for two to three weeks while blended numbers recover (Hustle Marketers).

Week 3: fix the conversion signal. Define the qualified lead in writing (see the JSON sketch above). Wire the CRM to the ad platforms: GCLID-based offline conversion imports on Google, Conversions API value events on Meta, Events API qualified-lead events on TikTok. If WhatsApp is the front door, score the conversation, not the opening message.

Week 4: restructure and replan. Rebuild the Meta account as the two-ad-set structure if existing-customer spend is high. Reallocate Google budget so brand defence is its own cheap campaign and PMax spends only on what Search cannot cover. Write the Ramadan calendar into the budget plan now. Then re-run the week 1 measurement and compare, on blended cost per qualified lead.

What to do next

The uncomfortable summary: these campaign types work, and in the UAE they fail in a specific, diagnosable, fixable way. The fix is mostly unglamorous plumbing: brand lists, exclusions, negative keywords and a conversion signal that means something. That plumbing is engineering work as much as marketing work, which is why we approach it as AI engineering: connecting the CRM to the bidding layer so the machine is told the truth about what a customer is worth. We do that work from Dubai, with our own team, alongside the media strategy (our AI engineering page covers the conversion-signal side; digital transformation covers the wider system and data mapping it sits on).

If you take one action from this piece: open your Performance Max search terms insights today and count how many of your conversions carry your own brand's name. Whatever that number is, that is the share of your budget that was buying demand you already owned.

https://www.azrty.com/blog/performance-max-advantage-and-smart-in-the-uae-keeping-ai-run-campaigns-from-buy

<a href="https://www.azrty.com/blog/performance-max-advantage-and-smart-in-the-uae-keeping-ai-run-campaigns-from-buy">Performance Max, Advantage+ and Smart+ in the UAE: keeping AI-run campaigns from buying demand you already own</a> (Azrty)