Are TikTok Spark Ads Worth It? What They Cost, Whether They Convert, and How They Compare to Meta and Google

Spark Ads are the format everybody recommends and almost nobody examines. The pitch is irresistible: take a video that already worked on TikTok, pay to show it to more people, and keep every like and comment it earns on the original post. TikTok publishes numbers to match, and they are enormous. What I wanted to know was whether those numbers survive contact with a sceptical reading, what the format actually costs once you include the minimum spend TikTok quietly requires, whether you can genuinely send someone to a product page from inside one, and how any of it compares to buying the same attention on Meta or Google. The short answer is that Spark Ads are a genuinely good format sold with a genuinely misleading statistic. I build small apps and run small budgets, so I use my own as the running examples: Repeat Recorder, Paint Vlix and Camera to Clipboard. There is a companion piece to this one on whether a viral Spark Ad ends up selling the trend instead of your product, which is the failure mode this article deliberately does not cover.

What a Spark Ad actually is: a paid boost of a real post, which is the only ad format where your budget permanently grows something you own

Almost every explanation of Spark Ads starts with the marketing description. The mechanical description is more useful, because the mechanics are where the advantages and the traps both live.

• It is an existing organic post, promoted. Not a video uploaded into the ads system, but a post that already lives on a TikTok profile, either yours or a creator's. The ad is a pointer to that post, not a copy of it.
• It runs under the original handle. The username, the profile photo, the caption and the comment section all come from the source post. Viewers see an account, not an advertiser.
• The engagement flows back to the original post, permanently. This is the part that genuinely has no equivalent elsewhere. TikTok's own documentation states that all views, comments, shares, likes and follows gained while the video is being boosted are attributed to the organic post. When the campaign ends, that engagement stays.
• A standard In-Feed ad is the opposite: a dark post. It exists only inside the ads account, has no home on any profile, and everything it earns disappears with the campaign. You rent the attention and hand it back.
• You are borrowing the post, not receiving a copy of the file. Which is why you cannot edit anything inside it, and why advertising somebody else's post needs their explicit permission. That permission is the authorisation code, explained just below.
• Which gives the format its real economic argument. Every other paid format converts money into temporary reach. This one converts money into temporary reach plus a permanent increase in the follower count, view count and comment volume of a post that keeps working after you stop paying. That residual is never in the cost per acquisition figure, and it is the strongest honest case for the format.

Every other ad format rents attention. A Spark Ad rents attention and keeps the receipts on an asset you own.

What the authorisation code actually is, and why you only need one for somebody else's post

This term gets used constantly in TikTok advertising advice and almost never explained, so here it is plainly. An authorisation code, also called a spark code, is permission to advertise a video you do not own.

• For your own posts there is no code at all. If your TikTok account is linked to your ads account, you simply pick the video from a list of your own posts. Nothing is generated, nothing expires, and none of the warnings later in this article apply. This is worth knowing because most guides describe the code as if it were a universal step, which puts people off a format they could use in two clicks.
• For a creator's post, the code is the entire handover. It is a string of characters the creator produces and sends you. Pasting it into your ads account unlocks that one specific video for you to advertise, and nothing else on their account.
• The creator generates it in four steps, inside their own TikTok app. Open the video, tap the three dots, tap Ad Settings, accept the advertising terms and switch on Ad authorization. They then choose a duration and tap Copy Code.
• Their account has to be a Professional account. On a personal account the Ad Settings option is not there at all, which is the usual reason a willing creator tells you they cannot find the button.
• The duration is chosen at that moment, from four options: 7, 30, 60 or 365 days. The creator picks it, not you, which is why it has to be part of the conversation rather than an afterthought.
• You paste it into Ads Manager under the TikTok posts tab, where TikTok also lets you authorise up to twenty video codes in one go if you are running several creators at once.
• And the clock starts when the code is generated, not when your campaign launches. That single detail is the source of the problem covered later: a 7 day code handed over during negotiation can be half expired before you have finished building the ad.

Yes, you can link out to your product, but the link is one of several exits and the others lead somewhere that is not you

This is the question I see asked most often and answered least precisely. The answer is yes, with a structural caveat that matters more than the yes does.

• The call-to-action button goes to your landing page. You attach a CTA and a destination URL when you build the ad, exactly as with any other TikTok ad format. Website traffic, app installs, lead generation and conversions are all available objectives.
• The caption link goes there too. If the ad carries a caption with a link, it points at the same destination.
• But the profile photo, the username and a left swipe do not. Per TikTok's help documentation, all three take the viewer to the profile page of whoever posted the video. If you sparked a creator's post, that is the creator's profile, not yours.
• The music or artist name goes to the sound page. Tapping it drops the viewer into a feed of every other video using that sound. That is a fully functional exit door out of your ad and into somebody else's content.
• The follow button follows the posting account. Again, the posting account, which on a creator spark is the creator.
• So the practical rule is that a Spark Ad has one door to your product and four doors elsewhere. This is not a defect, it is what makes the ad feel native instead of feeling like an ad, and the native feel is the entire point. But it means the video itself has to do the persuading, because a meaningful share of the people who engage will engage by going somewhere other than your landing page.

What TikTok ads actually cost: the floor on spend is higher than most people expect, and the CPM has been climbing about 16% a year

There is no separate price list for Spark Ads. They compete in the same auction as every other TikTok ad, so the cost question is really a TikTok cost question with one format-specific wrinkle at the end.

• The minimum daily campaign budget is above $50. TikTok's budget documentation is explicit: campaign budget must exceed $50, ad group budget must exceed $20. Since a campaign needs at least one ad group, roughly $50 a day is the practical floor.
• Lifetime budgets are worse than they look. The minimum lifetime ad group budget is the number of days multiplied by the daily minimum. TikTok's own worked example: a 31 day campaign needs at least $620 at the ad group level.
• Median CPM was $13.26 across 2025. That is from Triple Whale's full-year benchmark set, drawn from connected ecommerce accounts rather than from TikTok. It was up about 16% year over year from $11.43.
• Median cost per acquisition was $32.74, up about 8.6%, with conversion at 2.01% (down 6.2%) and return on ad spend at 2.21 (down 5.7%). Click-through rate went the other way, up about 13.7% to 1.77%.
• The spread by category is wide. In the same dataset, cost per acquisition ran from about $13.46 in pets and animals to about $31.25 in electronics, with most categories clustering between roughly $17 and $25.
• The direction of travel is the real headline. Costs up, conversion down, return down, clicks up. That is the signature of a maturing auction: more advertisers bidding for the same attention, and the attention getting slightly less valuable per unit as the novelty wears off. TikTok is no longer the arbitrage it was in 2021.
• The Spark-specific wrinkle is modest. TikTok has reported roughly a 4% lower CPM for Spark Ads than non-Spark In-Feed placements. Four percent is a discount, not a strategy.

The 134% claim: TikTok's own performance numbers are real measurements of a comparison that could not have gone any other way

Every Spark Ads guide on the internet repeats the same four statistics. They come from TikTok's own blog, and they are impressive.

• 134% higher completion rate than standard In-Feed ads.
• 157% higher 6-second view-through rate than standard In-Feed ads.
• 69% higher conversion rate and 37% lower cost per action with the profile landing page interface.
• Now here is the problem with all four. A Spark Ad is, by definition, made from a post that already existed and already performed well enough that somebody chose to promote it. The pool of Spark creative is pre-filtered for videos that human beings voluntarily watched. The pool of standard In-Feed creative contains every ad anybody made, including the ones that were bad from the first frame.
• That is survivorship bias, and it is enough to produce the entire gap on its own. If you compared "songs that already charted" against "all songs recorded", the charting group would show a spectacular listening advantage, and none of it would tell you that charting causes people to listen. The comparison cannot lose.
• The format almost certainly does add something real. Prior engagement is a genuine ranking signal, social proof on a visible like count is a genuine persuasion mechanism, and content shot by a creator for their own feed genuinely looks less like an ad than content shot for a brand. There are good mechanistic reasons to expect a positive format effect.
• But nobody has published the test that would size it. The question that matters is not whether Spark Ads beat In-Feed ads in aggregate. It is whether sparking this specific video beats running that same video as a standard In-Feed ad, same audience, same budget, same week. That is a trivially easy split test to run inside one account, and the absence of any published clean result is itself informative.

The statistic is not false. It just answers a question nobody asked: are videos that already worked better than videos that never did?

What you can do about it in one afternoon, for about a hundred dollars

• Run the test TikTok did not. Take one video. Build it two ways, as a Spark Ad from the organic post and as a standard In-Feed ad from the same file. Same targeting, same objective, same budget, same period.
• Judge on cost per outcome, not on completion rate. Completion rate is the metric most flattered by the format and least connected to money.
• Expect the gap to be smaller than 134% and larger than zero. If it is not, you have learned something specific to your account that no benchmark could have told you, which is the entire value of the exercise.

How TikTok compares to Meta on the one dataset that measures both the same way: cheaper to reach, cheaper to convert, worse at getting the click

Cross-platform comparisons are usually worthless because each number comes from a different source with a different definition. The exception is a dataset that reads both platforms from the same connected accounts in the same period. Triple Whale's 2025 set covers close to 35,000 brands and does exactly that.

• Reaching a thousand people cost less on TikTok: $13.26 against Meta's $14.19.
• Acquiring a customer cost less on TikTok: $32.74 against Meta's $38.19.
• Return on ad spend was better on TikTok: 2.21 against 1.86.
• Conversion rate was better on TikTok: 2.01% against 1.60%.
• Meta won the click: 2.19% click-through against TikTok's 1.77%.
• Meta is closing the gap from the wrong direction. Meta's CPM rose about 20% year over year while TikTok's rose about 16%, so both are getting more expensive and Meta faster. But Meta's conversion rate improved while TikTok's fell, which is the trend to watch.
• The honest reading is "comparable, with TikTok slightly ahead on cost". Not the tenfold difference the TikTok advocacy suggests, and not the disaster the sceptics suggest. Two large auctions selling similar attention at similar prices, one of them a little cheaper and a little less proven.

How TikTok compares to Google, where price is the wrong axis entirely: one platform sells attention and the other sells intent

Comparing a TikTok CPM to a Google Search CPC is a category error, and it is the error behind most "TikTok is cheaper than Google" claims. The two platforms are not selling the same thing.

• Google Search sells intent. Somebody typed the words. The click costs about $2.69 on average across industries and converts at about 2.81% for ecommerce, giving roughly $45 per conversion. Expensive traffic, cheap customers.
• Google Display sells attention, badly. Clicks around $0.63, ecommerce conversion around 0.59%, cost per conversion around $66. I went through why that inversion happens in the Google Display piece.
• TikTok sells attention, well. Nobody on TikTok asked for your product, but they are watching full screen with sound on, which is a fundamentally better attention product than a banner in a sidebar. TikTok's $32.74 median cost per acquisition beats both Google Display and Google Search on that dataset, which is genuinely notable.
• The catch is that attention platforms create demand and search platforms capture it. A TikTok campaign that works often shows up as a rise in branded search, direct traffic and app store searches, and only partly as clicks on the ad. If you are running both, some of what your Google Search line is being credited with was actually caused by TikTok.
• TikTok now sells a bit of intent too. TikTok Search Ads let you target keywords on the search results page, with TikTok recommending roughly 10 to 50 thematically grouped keywords per ad group. Volume is small compared to Google, and it costs more per click than In-Feed, but it is the intent half of the platform and it is new enough that competition is thin.
• So the real comparison is not TikTok versus Google. It is whether you have a demand problem or a capture problem. If nobody knows your product exists, Search has nothing to capture and TikTok is the better buy. If people are already searching for what you sell and you are not there, no amount of TikTok will fix that.

The measurement problem runs in both directions at once: last-click badly undercounts TikTok, and TikTok's own dashboard badly overcounts it

Both of these are true at the same time, which is why arguments about whether TikTok "works" go in circles. The two errors point in opposite directions and neither side is lying.

• Last-click undercounts, because TikTok is upstream. Somebody sees your video, does nothing, searches your name four days later and buys. Google Analytics credits organic search. TikTok caused it and gets nothing.
• TikTok's own reporting overcounts, because of the view-through window. The default attribution setting is 7-day click and 1-day view, and the view-through half means a conversion can be credited to your ad because the video appeared on the screen, with no interaction at all. On a platform where people scroll past dozens of videos a minute, that is a very large net.
• The self-reported and the analytics numbers will never match, and you should stop trying to make them. They are measuring different things. Expect TikTok's number to be several times your analytics number and treat neither as the truth.
• Incrementality is the only referee. TikTok's Conversion Lift Study is a proper holdout experiment: some of your audience is prevented from seeing the ads, and the difference is the causal effect. TikTok says it has run these with over 1,400 advertisers.
• The published lift results are worth reading with one eye open. TikTok's case studies show large gaps between last-click and measured lift, and a striking finding that a majority of incremental revenue in one study arrived in the weeks after the campaign ended. That is plausible for a demand-creation channel. It is also a marketing asset published by the platform whose product is being validated.
• The small-budget version of this test is a geographic holdout. Run TikTok in some regions and not in others for a month, and compare total revenue, not attributed revenue. It is crude, it needs enough volume to see through the noise, and it is still more honest than any dashboard.
• Set the view-through window to off while you are learning. You will see a much smaller and much more real number, and you can turn it back on later once you know what your true baseline looks like.

The gotchas that quietly kill Spark campaigns: an expiring code, a caption you can never edit, a comment section you do not control, and music rights that do not transfer

These are not edge cases. Each one has a specific mechanism and each one is invisible until it costs you something.

• The authorisation code expires on a timer, and your ads stop dead. This one applies to creator posts only, since your own posts need no code. Codes are issued for 7, 30, 60 or 365 days, counted from the moment the creator generated it rather than from your launch date. When the window closes the code goes inactive, any live ads using it stop running, and it cannot be reactivated. You need a brand new code from the creator and a rebuilt ad. Nobody warns you in advance, so the first symptom is usually a campaign that flatlined overnight.
• Set the longest duration you can get, every time. There is no cost to a 365 day code and no benefit to a short one, other than the creator's comfort. Ask for the long one up front, because renegotiating after a campaign stops is worse.
• You cannot edit the caption after the ad is created. Finalise the copy first. Anything wrong in it is permanent for that ad's life.
• You cannot edit the video at all. No trimming, no reframing, no adding a CTA card. What the creator posted is what you are buying, which is the price of authenticity.
• The comment section comes along, and it is live. You are paying to distribute other people's replies as well as your video. On a good post that is the strongest social proof money can buy. On a post with a sceptical top comment, you are paying to amplify an objection to your own product to a much larger audience than ever saw it organically.
• Read the top comments before you spend, and re-read them daily after. This is the single highest-value five minutes in running a Spark Ad, and almost nobody does it.
• Music rights do not transfer from organic to paid. A trending consumer sound is licensed for personal posts, not necessarily for advertising. Boosting the post does not launder the licence. TikTok maintains a separate Commercial Music Library of pre-cleared tracks precisely because of this, and it is why some perfectly good organic posts simply cannot be sparked.
• You cannot delete the organic post while it is authorised. The source has to stay up. Un-authorise first, then delete.
• Videos have a ten minute ceiling. Rarely a constraint, occasionally fatal for long-form content.

Spark Ads, whitelisting, dark posts and Promote: four different things that get discussed as if they were one

Most of the confusion in TikTok advertising advice comes from these four being used interchangeably. They are not interchangeable, and the differences decide who controls what. Click any column header to sort (best first; click again to reverse).

• Promote is the trap for small budgets. It looks like the cheap easy version of a Spark Ad and it is really a stripped-down toy: almost no targeting, no proper objectives, no useful reporting. If you are spending enough to care about the result, use Ads Manager.
• Whitelisting is Spark Ads with the training wheels off. You get the creator's account access and can build entirely new ads under their handle, including videos they never posted. More power, more paperwork, more trust required, and the follower growth is theirs.
• Dark posts still have a job. If you want to run fifteen creative variants without publishing fifteen posts to a profile, that is exactly what standard In-Feed is for.

The five realistic ways to buy this attention, compared on what actually differs rather than on headline price

Here is the whole landscape in one place. Click any column header to sort (best first; click again to reverse). Sorting by "Buyer intent" and then by "Cost per customer" shows the trade the entire industry runs on.

No row wins outright, which is the point of laying them side by side. TikTok buys the cheapest customers and the least intent. Google Search buys the dearest clicks and the only certainty about what the person wanted. Spark Ads are the only row where the money leaves something behind, and the only row where the reporting is least trustworthy by default.

The platform is changing underneath the format: automation is eating the controls, and the US business now sits inside a different company

Two structural changes matter more to a 2026 media plan than any creative tactic, and both are recent enough that most guides have not caught up.

• GMV Max replaced the shopping ad formats outright. From July 2025, TikTok consolidated Product, LIVE and Video Shopping Ads into a single automated campaign type optimising for gross merchandise value. For TikTok Shop destinations the legacy formats can no longer be created.
• And it took the audience controls with it. For shop sales objectives you can no longer choose age, gender or interests. The system decides who sees the ad. Whether you like it or not, that reallocates your job from targeting to creative.
• Smart+ is the same idea for everything else. A single automated buying flow across traffic, leads, app installs and web conversions, with the amount of automation dialable rather than mandatory.
• Which makes Spark Ads more important, not less. When targeting levers disappear, the creative asset becomes the only lever left, and a Spark Ad is the highest-quality creative asset most small advertisers can obtain, because it was validated by an audience before you paid for it.
• The US business is now a separate joint venture. The deal closed on 22 January 2026. TikTok USDS Joint Venture LLC is roughly 80% owned by US investors, with Oracle, Silver Lake and MGX each holding about 15% and ByteDance retaining 19.9%. Oracle is the security partner and the US recommendation algorithm is operated independently and retrained on US data.
• The practical effect for advertisers is stability plus uncertainty. The existential ban risk that made planners avoid TikTok for two years is largely resolved. In exchange, the algorithm serving US users is being retrained, so historical performance is a weaker guide to future performance than it usually would be. Re-test rather than assume.

So are Spark Ads cost efficient? Yes, with the important caveat that the format is not what makes them work

Here is the whole thing as a working method rather than a verdict.

• Spark Ads are cost efficient because of what they are made of, not because of what they are. The format adds a modest CPM discount and a genuine social proof effect. The large advantage comes from the creative selection: you are promoting something an audience already voted for. Any process that reliably picks good creative would capture most of the same gain.
• Which gives you the actual strategy: post organically to generate candidates, then spark the winners. Treat organic posting as a cheap testing lab for paid creative. This is the single highest-return workflow on the platform and it costs nothing but time.
• Budget realistically or not at all. About $50 a day is the floor, and you need enough days for the system to exit its learning phase. A two day, $100 experiment tells you nothing except that you spent $100.
• Expect roughly $30 to $35 per customer on ecommerce, and treat that as a starting hypothesis. Category matters more than platform: the same campaign structure that lands at $13 in pets can land at $31 in electronics.
• Ask for the long authorisation code, finalise the caption before you launch, and read the top comments daily. Three habits that prevent most of the ways Spark campaigns die.
• Turn the view-through window off while you are learning what is real. Then run a holdout, geographic or otherwise, before you make any decision bigger than the budget you have already spent.
• Compare against Meta on your own account, not on benchmarks. The independent data says TikTok is slightly cheaper on both cost per thousand and cost per customer. Slightly is not enough to be true for everyone.
• Do not use it to capture demand that already exists. If people are typing your category into a search box, that budget belongs on Google Search. TikTok is for the demand that has not been created yet.
• And read the failure mode before you write the creative. The hook that makes a Spark Ad cheap is the same hook that can make people remember the joke and forget the product, which is the vampire effect, and it has a TikTok-specific version.

A Spark Ad is not a clever ad format. It is a distribution deal with a video that already earned its audience, and the reason it works is that somebody else already did the hard part.

The summary, in one glance

• A Spark Ad boosts a real organic post, runs under a real handle, and permanently banks its likes, comments, shares and follows on the source post. No other paid format leaves anything behind.
• You can link out to your product, via a call-to-action button and the caption, but the profile tap, the left swipe, the follow button and the sound name all lead somewhere else.
• The real floor on spend is about $50 a day, with a $50 campaign minimum and a $20 ad group minimum, and lifetime budgets multiply that by the number of days.
• Median TikTok cost was $13.26 per thousand and $32.74 per customer in 2025, with CPM up 16% year over year and both conversion and return sliding. The arbitrage era is over.
• TikTok's 134% and 157% claims are survivorship bias. Spark creative is pre-filtered for videos that already won; In-Feed creative is not. The format effect is probably real and certainly smaller.
• The test TikTok never published costs about a hundred dollars. Same video, sparked versus standard, same audience and budget, judged on cost per outcome.
• TikTok beat Meta on the same dataset, $13.26 against $14.19 per thousand, $32.74 against $38.19 per customer, 2.21 against 1.86 return. Meta won click-through, 2.19% against 1.77%.
• Against Google it is attention versus intent. Google Search costs about $2.69 a click and converts about 2.81%, because the person typed the words. TikTok is for creating the demand that Search later captures.
• Both measurement errors are real at once. Last-click undercounts TikTok because it works upstream; TikTok's dashboard overcounts because view-through attribution is on by default.
• Four gotchas kill campaigns: the authorisation code expires with no warning, the caption is uneditable, the comment section is live and amplified, and music licensed for organic posts is not automatically licensed for ads.
• Promote is not a cheap Spark Ad, it is a toy. Whitelisting is Spark Ads with more power and more paperwork. Dark posts are for volume testing.
• Automation is removing the targeting levers, with GMV Max already the only route for TikTok Shop and audience controls gone for shop objectives, which makes creative the only lever left.
• The US business is now a separate joint venture as of January 2026, which removes the ban risk and adds algorithm-retraining risk in its place.