Junk leads from Dubai ad campaigns: causes and the filters that fix them

Most junk leads from Dubai campaigns are not a bot problem. Four defaults do most of the damage: Google puts new Search campaigns on search partners and the Display Network, Google's location option ships as "Presence or interest" rather than presence, Meta's instant form defaults to a one-tap screen with pre-filled contact details, and the website form has no submission control beyond hope. Change those four and the sales manager stops dialling dead numbers, usually within a week.

Written by Artur Gall, CEO of SkyLight Marketing. We run paid media out of Dubai, and our named client work sits mostly in fashion, beauty and jewellery e-commerce (Fabiana Filippi, DSQ Cosmetics, Toktam Jewelry, Rayhaan). The lead-quality war is a service-business problem, so everything below comes from service accounts and from our own inbound, not from a vendor whitepaper.

What a junk lead actually costs you

I am not going to quote you a percentage. The "40 to 50 percent of UAE leads are junk" figure circulates on agency blogs without a survey behind it, and nobody publishes a methodology I would defend in a client meeting. Calculate your own instead, because your number is the only one that changes decisions.

Take last month. Count every enquiry the campaign produced. Then count how many your team reached on the first or second attempt. The gap is your junk rate. Now push it through the cost:

Line	Where to get it
Ad spend last month	Google Ads and Meta billing
Enquiries received	CRM or inbox count
Reachable enquiries	Anyone who answered or replied
Real cost per reachable enquiry	Spend divided by reachable enquiries

The second number is the one to put in front of whoever approves budget. If half your enquiries never answer, your true cost per contactable person is double what the platform reports, and every bid decision the algorithm made last month was based on the wrong signal. That is the real damage. The wasted staff hours are annoying; the mistrained bidding is expensive.

For channel-by-channel bands to sanity-check your figure against, we keep those in the UAE cost per lead benchmarks piece rather than repeating them here.

Where the bad enquiries come from, source by source

Junk arrives in recognisable flavours, and each one points at a different setting.

What you see	Usual source	What to change
Number rings once, then disconnected	Meta instant form autofill of an old profile number	Form type and phone question
Person has no idea what your business is	Display Network or search partners inside a Search campaign	Networks setting
Enquiry from outside the UAE	"Presence or interest" location option	Location options
Same person submits four times	No deduplication on the form or CRM	Submission logic
Gibberish name, valid-looking email	Website form with no bot control	Honeypot plus timing check
Right country, wrong service entirely	Broad match with no negative list	Keywords and search terms

Two of these are worth naming carefully. Accidental submissions on Meta are genuinely common because the form arrives pre-filled from the profile, so a thumb brushing the screen produces a complete lead record with a phone number that may be years old. And a Search campaign that quietly runs on the Display Network produces enquiries from people who tapped a banner inside a game, which is why they sound baffled on the phone.

Run this table against your last fifty enquiries before touching any settings. The distribution tells you which fix to do first.

The Meta form settings that change who submits

Open the instant form and switch the form type from "More volume" to "Higher intent". The higher-intent form adds a review screen before submission, so the person confirms their details instead of one-tapping through. It costs you volume. That is the point, and it is the single highest-leverage change in the whole Meta setup.

Two working notes on it. Higher-intent forms are delivered on Facebook and Instagram mobile feeds rather than across every placement, so check your placement mix before you blame the change for a delivery drop. And the review screen filters accidents, not liars, so pair it with the next step.

Replace the pre-filled phone question with a manually typed one. Meta lets you set a phone question that does not autofill, which forces the person to type a number they currently use rather than accepting whatever sits in their profile from a SIM they abandoned in 2019. Add one custom multiple-choice question that a browser will not bother answering, and make it commercially meaningful rather than a quiz: budget band, timeline, emirate, or which service they want. A single well-chosen question does more filtering than three vague ones, because each extra screen also loses people who would have bought.

Set the form language to match the ad. If the creative runs in Arabic and the form loads in English, you get abandonments and half-completed records that read like spam but are not.

If you want this configured properly across an account, that sits inside our paid social work.

Four Google Ads settings and one keyword habit

Start in campaign settings, under Networks. Uncheck "Include Google search partners" and uncheck the Display Network. New Search campaigns are created with both switched on, and neither pauses your ads or resets anything when you turn them off. Display traffic inside a Search campaign is the most reliable junk generator in the platform.

Second, open Location options, which sits collapsed under the locations panel. The default is "Presence or interest: people in, regularly in, or who've shown interest in your targeted locations". For a Dubai service business that means someone in Manila researching a move to the UAE counts as targetable. Switch it to "Presence: people in or regularly in your targeted locations". Real estate and education accounts sometimes lose real volume here, so change it, then check search terms and enquiry quality for two weeks before you decide.

Third, build a negative keyword list that reflects your actual business rather than a generic template. The categories that earn their place in the UAE: job seekers (jobs, vacancy, hiring, salary, career), price-only researchers who will never buy your tier (free, cheap, DIY, how to make), competitor names you do not want to pay for, and course or training variants if you sell a service rather than teaching it. Pull it from the search terms report every Monday for the first month. Templates you copy from a blog will not contain the phrase that is actually draining your budget.

Fourth, treat broad match as a decision, not a default. Broad match with a well-fed conversion signal behaves reasonably. Broad match with no conversion tracking, or tracking that fires on every form submission including the fake ones, is how an account teaches itself to buy the wrong people. Which brings us to the part most agencies skip.

Our Google Ads management starts with exactly this audit sequence on every new account, because it is faster than rewriting creative.

Verifying the number, and why WhatsApp beats a callback

A UAE mobile is nine digits after the country code: +971, then a 5, then a two-digit prefix, then seven digits. The live mobile prefixes as of 2026 are 50, 52, 53, 54, 55, 56 and 58. A format check that rejects anything else removes a surprising share of nonsense before it reaches a human:

^(?:\+?971|0)?5[0234568]\d{7}$

One myth to kill while you are in there. The prefix used to tell you the operator (050, 054 and 056 for Etisalat, 052, 055 and 058 for du, 053 for Virgin Mobile, with 058 shared between du and Virgin), but number portability means a 050 number can sit on any network today. Validate the format and whether the number is reachable. Do not try to infer the carrier, and do not let a developer build a rule on it.

On the website form, two controls do most of the work and neither annoys a human. Add a hidden field with a plausible name that real users never see and bots fill in, and reject any submission where it contains text. Add a timestamp check and reject submissions completed in under three seconds, because nobody types their name and number that fast. If you also run reCAPTCHA v3, remember it returns a score rather than a pass or fail, and the default threshold of 0.5 is a starting point you should tune against your own traffic rather than treat as gospel.

Then move the intake to WhatsApp. In the UAE, a person who taps through to WhatsApp has demonstrated they own the number, which is a stronger signal than any validation script produces. We answer our own inbound on WhatsApp within minutes, not hours, and speed matters more than script quality at that stage. We covered the choice between calls, forms and WhatsApp in detail in the piece on call assets and lead form assets, so pick your channel there and come back here to filter it.

Teaching the platform which leads were real

Filters clean your inbox. Feedback cleans the algorithm, and that is the part almost nobody does.

On Google, import the outcome back into the account. Capture the GCLID when the form is submitted, store it against the record, and upload a conversion when the lead is qualified, with a value that reflects reality: a small value for a qualified enquiry, a larger one for a closed deal. Enhanced conversions for leads is the current version of this, using hashed user-provided data alongside the GCLID. Note the calendar: from June 2026 these uploads move to the Data Manager API and are blocked in the Google Ads API, so if your setup was built a year ago, check who owns that integration before it stops silently.

On Meta, the equivalent is Conversions API for CRM. You send lead stage events back (lead created, qualified, converted) keyed on the Meta lead ID, then switch the campaign performance goal from maximising leads to the conversion leads goal. Meta needs that CRM feed to optimise for quality rather than count. One practical constraint: Meta wants the stage you optimise for to convert somewhere between roughly 1 and 40 percent of leads, with real volume behind it. A business that closes one deal in two hundred leads falls below that floor and won't give the system enough to learn from, and "lead created" itself converts at effectively 100 percent, so it doesn't qualify either. Pick a stage in between, such as qualified or quoted.

This is the highest-return work on the list and the slowest to show results. Give it four to six weeks before judging it.

Language and geography in a UAE audience

Ad language changes who answers, not just who reads. English campaigns in Dubai reach a broad expat population; Arabic campaigns reach a different buyer with different expectations about how quickly you respond and who calls whom. Running one ad set in both languages blends two audiences into one bid and one set of learnings, and the enquiries look inconsistent as a result. Split them, then compare the reachable-enquiry rate rather than the raw count.

Geography deserves the same discipline. Targeting "United Arab Emirates" when you serve Dubai and Sharjah adds enquiries you cannot service, and an unserviceable enquiry is junk even when the person is entirely real. Radius targeting around the areas you actually cover beats country targeting for any business with a van, a clinic or a showroom.

Split by language for a month, then decide with data instead of intuition.

What to look at instead of cost per lead

Cost per lead rewards whoever buys the cheapest clicks. Four metrics that do not:

Metric	How to calculate	Why it matters
Contact rate	Reached divided by total enquiries	Catches fake and dead numbers
Cost per reachable enquiry	Spend divided by reached	The honest version of CPL
Cost per qualified enquiry	Spend divided by qualified	Catches wrong-service traffic
Time to first response	Minutes from submission to first reply	The one you control today

Track contact rate weekly per campaign, not per account. Junk clusters in one campaign or one ad set almost every time, and an account-level average hides it perfectly.

If you want someone to run this audit across your Google and Meta accounts and hand you the settings list with the search terms report attached, tell us what you are running and we will look at it.

FAQ

Why do I get fake leads from Facebook and Instagram ads in Dubai? Usually accidental submissions rather than bots. Meta's instant form pre-fills contact details from the profile, so a mistap produces a complete record with an outdated phone number. Switching the form type to "Higher intent" adds a review screen, and replacing the pre-filled phone question with a manually typed one removes most of the rest.

How do I stop spam submissions on my website form? Add a hidden honeypot field that bots fill and humans never see, reject submissions completed in under three seconds, and deduplicate on phone number within a 24-hour window. If you use reCAPTCHA v3, tune the score threshold against your own traffic rather than leaving it at the default 0.5.

Which negative keywords cut junk in UAE Google Ads campaigns? Job-seeker terms (jobs, vacancy, hiring, salary), free and DIY variants, course and training terms if you do not teach, and competitor brand names you do not want to pay for. Build the list from your own search terms report weekly for the first month instead of copying a generic list.

How do I check a UAE phone number is real? Validate the format first: +971 followed by 5, then one of the prefixes 50, 52, 53, 54, 55, 56 or 58, then seven digits. Then test reachability on WhatsApp. The prefix does not tell you the operator any more because of number portability, so do not build logic on that.

Is an instant form or a landing page better in Dubai? Instant forms give volume and lower cost per enquiry; landing pages give context and better-informed enquiries. If your team is drowning in unreachable numbers, move to a landing page with a WhatsApp button, or keep the instant form and switch it to higher intent. Test both against contact rate, not cost per lead.

Why does my CRM show 40 percent of leads as "no answer"? Three causes in order of likelihood: outdated autofilled numbers from instant forms, response delays measured in hours rather than minutes, and calls placed from an unknown landline that nobody in the UAE picks up. Fix response time first, because it costs nothing and it is usually the biggest single factor.

Does ad language affect lead quality in the UAE? Yes. Arabic and English campaigns reach different buyers with different response expectations. Running both in one ad set blends the signals and produces inconsistent enquiries. Split them and compare reachable-enquiry rates.

What is a normal cost per lead in Dubai? It varies by industry too widely for a single number to be useful, and the figure matters less than cost per reachable enquiry. We publish channel bands in our UAE cost per lead benchmarks article; use them as a sanity check on your own arithmetic rather than a target.