Per our own views-to-likes benchmark of 5,375 creators, an engagement ratio outside the observed band is the cheapest tell that a launch was amplified.
FORKOFF Research · RADAR · Data Study · 2026-08-25
FORKOFF RADAR read 30 public X product launches by the views-per-like method. Two in three, 20 of 30 or about 67%, show a bought-amplification signature, and only 33% read cleanly organic. Every figure is read from public data with the published RADAR method.
Paying for distribution is legal, common, and often a sound launch tactic. This study measures how reach was built, not whether anyone was honest. It is reported at the aggregate level, and the only launches named are the ones that read cleanly organic. This page is the method and dataset behind the read. For the current-state numbers on their own, see the companion Launch Authenticity Report.
Of the 30 public X product launches FORKOFF RADAR scored on the views-per-like method, 20 of 30 (about 67%) carry a distribution-amplified signature and 10 (33%) read cleanly organic. Organic reach on X tops out near 500 views per like; above that the reach ran ahead of the audience, and above roughly 5,000 the ratio reaches the fraud-tier band. Paying for distribution is legal and common, so this study measures how reach was built, not whether anyone was honest.
Likes are the cheapest genuine signal a real audience gives a post, so the ratio of views to likes is a clean tell for how reach was built. RADAR divides each launch post's public view count by its public like count, then reads the ratio against a ladder calibrated on real launches. The organic ceiling sits near 500 views per like. The heavy distribution band begins near 2,000. The fraud-tier band begins near 5,000, where the impression count and the engaged audience come apart.
To keep a light read from being mislabeled heavy, RADAR also weighs the costliest actions to manufacture, replies and quote posts, against the reach. A launch with a thick written conversation under it earns a lighter read at the same ratio than one where the reach stands alone.
Under ~500:1
Organic
Views and likes climb together. The reach reads as earned, with no amplification signature.
~500 to ~2,000:1
Paid-lite
The reach runs a modest step ahead of the likes. A light amplification layer on a launch that was still connecting.
~2,000 to ~5,000:1
Paid
The reach runs well ahead of the engagement, in the heavy distribution band.
Over ~5,000:1
Botted-signature
A fraud-tier ratio where the impression count and the audience that engaged come fully apart.
The study reads 30 public product launch posts on X, published between October 2025 and June 2026, and scores each one on the ratio of its public view count to its public like count. Per our own record of those posts the corpus carries 110.4M+ views, 145,493 likes, 20,088 reposts, 23,152 replies, and 9,642 quote posts. Nothing in it is modelled.
Three rules decide whether a launch is in. The post has to be the launch itself, put out by the product account or the founder, not a recap thread or a press pickup. Every figure has to be a matured one, re-pulled after the launch settled, because a like count read in the first hour is still climbing while the impressions have already run, and reading the two together manufactures a views-per-like gap that says nothing about how the reach was built. And where a metric is not published, the launch stays out rather than being estimated in. A launch with a great story and no public like count is a launch this method has nothing to divide.
The 30 launches span 6 market verticals, from AI and dev tools through fintech, crypto, marketing, and consumer and health, which is what keeps the finding from being a single-niche curiosity. The per-vertical breakdown is on the RADAR coverage grid. Every launch in the corpus also keeps its own page with the source post cited and the reading shown in full, indexed on the RADAR hub, and the clean set is browsable in the verified organic directory.
A like is the cheapest genuine action a real person takes on a post, and the same feed placement that delivers a view is what makes the like easy to give. So when views and likes stop climbing together, the extra views arrived by a route the audience did not take. That is the whole idea behind the ratio, and it is why RADAR divides one public number by the other rather than reading either on its own.
Across this corpus the ratio runs from 98 views per like at the tightest to 12,075 at the widest, a spread of about 123 times on a signal whose two halves are both published on the post. The corpus median sits at 821, which is already above the roughly 500 organic ceiling. The middle of this set of launches is not a clean read, and that is the finding underneath the headline share.
What the ratio does not settle is how a lift was delivered, which is why it is never the only input. Replies and quote posts are the costliest actions to manufacture, because each one is an original written post from an account willing to attach its name, so a wide ratio sitting under a thick written conversation reads differently from the same ratio with nothing under it. The thresholds, the calibration, and the confidence rules are written out on the RADAR methodology page, and the ratio itself is benchmarked against a wider creator sample in our views-to-likes benchmark.
Across 30 public product launches RADAR read by the views-per-like method, 20 of 30 (67%) carry a distribution-amplified signature and 10 read cleanly organic. Bar length is the share of the 30-launch corpus. Paying for distribution is legal and common; this measures how reach was built, not whether anyone was honest.
reads earned
Views-per-like at or near the roughly 500 organic ceiling. RADAR finds no distribution-amplified signature.
light lift
Reach runs a modest step ahead of the engagement, a light amplification layer on a launch that was connecting.
heavy lift
Reach runs well ahead of the likes, above the roughly 2,000 line where the heavy distribution band begins.
fraud-tier ratio
A fraud-tier views-per-like ratio, above roughly 5,000, where the impression count and the audience come apart.
Source: the RADAR reference-launch corpus, 30 public launch posts scored on the views-per-like method against the roughly 500 organic ceiling and the roughly 5,000 fraud-tier band. Counts are computed directly from the underlying data, not estimated. The same forensic method audits every view on a FORKOFF viral launch engagement.
Split on the published ladder alone, 9 of the 30 launches sit at or under the roughly 500 organic ceiling, 13 sit in the light-lift band between 500 and 2,000, 6 sit in the heavy distribution band between 2,000 and 5,000, and 2 sit above 5,000 in the fraud-tier range. The table below is that split with the reach, the engagement, and the evidence behind each band, every figure computed from the corpus rather than typed.
Read the reach column alongside the launch count. The 21 launches sitting above the organic ceiling carry 85.3M views between them on 74,255 likes, which is the part of this corpus a founder is most likely to be shown in a launch report and least likely to be shown the engagement behind. That is the split by ratio, which counts 1 launch more than the 20 carrying an amplified reading, for the reason the disagreement section below sets out.
| Ratio band | Views per like | Launches | Share of corpus | Combined reach | Combined likes | Observed ratios | Full trace |
|---|---|---|---|---|---|---|---|
| Organic range | 500 views per like and under | 9 | 30% | 25.1M | 71,238 | 98 to 498 | 6 of 9 |
| Light lift | 500 to 2,000 | 13 | 43% | 47.2M | 61,100 | 556 to 1,953 | 6 of 13 |
| Heavy distribution | 2,000 to 5,000 | 6 | 20% | 31.9M | 12,480 | 2,088 to 4,410 | 5 of 6 |
| Fraud-tier range | over 5,000 | 2 | 7% | 6.2M | 675 | 5,824 to 12,075 | 1 of 2 |
| All bands | median 821 | 30 | 100% | 110.4M | 145,493 | 98 to 12,075 | 18 of 30 |
A ratio band is not a verdict. This split is mechanical, computed from the published ladder and nothing else, while the reading on a launch also weighs the written engagement layers and whether a full forensic trace exists. The two disagree on 1 of the 30 launches, which the next section takes apart rather than smoothing over.
The sharpest result in this corpus is not the ratio, it is how little extra audience the extra reach bought. The 10 launches that read organic carry 28,882,557 views and 72,619 likes between them. The 20 carrying an amplified signature carry 81,536,075 views, about 2.8 times the reach, and 72,874 likes, which is within half a percent of the same number.
The written layer separates them further. Counting replies and quote posts together, per our own reading of the corpus the organic set produced 14,037 written actions on its reach, about 486 per million views, against 18,757 for the amplified set, about 230 per million. Roughly twice the written conversation per view, on a third of the impressions. Written actions are the ones that cost an account something to give, which is why the gap widens exactly there.
The practical version for a founder is short. Reach and audience are two different purchases, and an impression count on its own cannot tell you which one you got. That is the reason FORKOFF reports a guaranteed view tier with the engagement behind it rather than an impression number on a slide, and the reason the same read runs on every launch video it delivers.
These are public launch posts, not FORKOFF client work. RADAR read each one by the views-per-like method and found no distribution-amplified signature, so the reach reads as earned. Bar length is the public view count on the launch post.
Launches RADAR read as amplified are deliberately excluded from this board. Every launch below links to its cited source post and full RADAR reading.
views-per-like: 312· organic
views-per-like: 492· organic
views-per-like: 396· organic
views-per-like: 211· organic
views-per-like: 445· organic
views-per-like: 498· organic
Source: the RADAR reference-launch set, public post metrics plus the views-per-like read. 6 organic-verdict launches shown, ranked by public reach. The same forensic method audits every view on a FORKOFF viral launch engagement.
RADAR names the organic launches because a clean read is a positive finding. All but 1 of these sat at or under the roughly 500 organic ceiling, where the engagement kept pace with the reach, so RADAR found no distribution-amplified signature. The launches that read amplified are reported only in the aggregate, never named. The list is drawn from the same safe set the directory and the leaderboard use, so it can never disagree with the 10 in the count above.
9 launches sit at or under the organic ceiling on the ratio. 10 read organic. The difference is 1 launch, MetaMask Money Account, which posted a ratio well above the ceiling and still reads organic, because the reading weighs the written engagement and the forensic trace rather than the ratio alone. RADAR publishes that disagreement rather than rounding a ratio into a verdict.
The general rule behind it is that the ratio narrows the question and the trace answers it. A wide ratio says the impression count and the engaged audience separated; it does not say who delivered the impressions or why. Where the trace shows the reach arriving on a curve the engagement followed, and where the written layers came in at a real volume, the reading moves toward organic even though the headline number looks heavy.
In this corpus the disagreement runs one way only. Every launch at or under the ceiling also reads organic, so on these 30 launches a tight ratio has never been contradicted by the trace. That is a fact about this corpus, not a rule, and a tight ratio has never been treated as an automatic clearance: the check still runs. Each launch keeps its own page with the components it was read on, so a reader can see which signal did the work.
RADAR labels two kinds of reading apart. 18 of the 30 launches carry a full forensic trace, and 12 are reconstructed from the published ratio and the engagement shape alone. A reconstructed launch can never earn a high-confidence amplified verdict, however wide its ratio looks, because the evidence needed to describe how a lift was delivered was never collected.
That rule bites hardest exactly where a stronger claim would be most tempting. Of the 20 launches carrying an amplified signature, 11 rest on a verified trace and 9 are reconstructed and therefore published as hedged, lower-confidence readings. In the fraud-tier band at the top of the ladder, only 1 of the 2 launches carries a trace, so the widest ratios in the corpus are not the most confident readings in it.
The same split produces the two published letters on the favorable side. 7 of the 10 organic reads carry an A for a verified organic reading and 3 carry a B where the organic read is reconstructed. Only organic reads ever carry a public letter. The ranked version of that safe set is the RADAR leaderboard.
A real public launch, read on the method FORKOFF uses to audit every view. Below the roughly 500-to-1 organic ceiling, the reach reads as earned. Above it flags a distribution-amplified signature. This dial is a public launch, not a FORKOFF client.
The launch on the dial
OpenAI's Jalapeño AI chip
@OpenAI · 2026-06-24
At 312 views per like the engagement keeps pace with the reach, inside the organic ceiling, so RADAR finds no distribution-amplified signature. This is the same read FORKOFF runs on every view of a guaranteed viral launch.
See the full RADAR reading →Buying or coordinating distribution on X is legal, common, and frequently a sound way to launch. An amplified signature is a statement about how an impression count was built, not about whether a product is real, a founder honest, or a metric faked. Every product in this corpus is a real product shipped by a real team, and the study says so on purpose rather than leaving it to be inferred.
That is why the reporting is asymmetric. The 10 organic reads are named, linked, and graded, because a clean read is a positive finding and the founder behind it should get it. The 20 amplified reads appear only in the aggregate, in a share and a band count, and no row anywhere on this page identifies one of them. That policy is not a legal disclaimer bolted on at the end; it is what the page is built around.
RADAR also reads a signature and never an intent. The method cannot separate a paid distribution partner from a genuine burst of attention delivered by a few very large accounts, and it does not try. It can say that the reach and the audience came apart, by how much, and with what confidence. Anything past that would be a guess wearing a number.
30 launches is a small corpus and a selected one. A launch enters it because its metrics were public and complete and its product name unambiguous, which is not the same as a random draw from every launch on X. So the 67% figure describes these 30 launches honestly and should not be read as a population estimate for the platform.
Three more limits are worth stating. A view is an impression and not a person, so every reach figure here counts deliveries rather than humans. The per-vertical counts run from 2 to 9 launches, which is too thin for any single vertical to carry a reliable rate on its own, and the breadth grid on the coverage page should be read as coverage rather than as six separate findings. And 10 of the 30 launches published at the top of the hour, of which 1 reads organic, which is a scheduling shape worth noticing on a sample of 10 and nowhere near evidence on its own.
The corpus grows by re-reading, not by rounding. Every launch keeps its own page with the source post cited, so a founder who thinks a reading is wrong can point at the post and the metrics and ask for a re-read, and a corrected reading changes this study on the next build. Every count, share, and band on this page is computed from the corpus at build time, which is the only reason the page can be trusted to still be true a month from now.
The same views-per-like read in this study runs on every view FORKOFF delivers on a launch. When FORKOFF guarantees a view tier on a viral launch, it hits that tier through organic distribution and a make-good if a number falls short, never through bought amplification that a RADAR read would flag. The point of measuring the market this way is to hold FORKOFF's own work to it.
That distribution engine has already processed 5B+ views across FORKOFF's clipping network, so the reach on a FORKOFF launch is reach a real audience saw, and the report you get back carries the same read this study runs on the public corpus.
In FORKOFF RADAR's 30-launch forensic corpus of public X product launches, 10 of 30 (about 33%) read cleanly organic. The other 20 (about 67%) carry a distribution-amplified signature, meaning the reach ran ahead of the audience that engaged with it. Paying for distribution is legal and common, so this is a measure of how reach was built, not an accusation against any launch.
Organic reach on X tops out near 500 views per like. Below that, views and likes climb together and the reach reads as earned. Between roughly 500 and 2,000 views per like the reach runs a step ahead of the audience, which RADAR reads as a light amplification layer. Above roughly 5,000 views per like the ratio reaches the fraud-tier band, where the impression count and the engaged audience have come apart. For scale, the ratio across this corpus runs from 98 to 12,075 with a median of 821.
RADAR divides a launch post's public view count by its public like count to get a views-per-like ratio, then reads that ratio against a calibrated ladder: a roughly 500 organic ceiling, a roughly 2,000 heavy-distribution line, and a roughly 5,000 fraud-tier band. Likes are the cheapest genuine signal to earn from a real audience, so when views vastly outnumber likes the reach was built by distribution rather than by the audience. Replies and quote posts, the costliest actions to manufacture, are checked to keep a light read from being mislabeled heavy.
The 10 launches that read cleanly organic are OpenAI's Jalapeño AI chip, Cursor for iOS, Koji, MetaMask Money Account, Short Video Overviews (NotebookLM), Contra Payments, Parker, Seed Audio 1.0, Browserbase Agents, Jupiter Wallet biometrics. All but 1 sat at or under the roughly 500 organic ceiling, where the engagement kept pace with the reach, so RADAR found no distribution-amplified signature. RADAR publishes the organic launches by name because a clean read is a positive finding.
No. Buying or coordinating distribution is legal, common, and often a sound launch tactic. An amplified signature describes how the reach was built, that the impression count ran ahead of the audience that engaged, and says nothing about whether the product is real or the founder honest. The products in this corpus are real products by real founders. RADAR reads a signature, never an intent.
The 67% figure is the current, reproducible reading of the 30-launch corpus: 20 of 30 launches carry a distribution-amplified signature. It is computed directly from the underlying data on every build, so the number on the page always matches the corpus. Earlier, higher figures were drawn from a smaller sample and are not carried forward.
Take the launch post's views and divide by its likes. If the result is under about 500, the reach kept pace with the audience and reads organic. If it is well above 500, and especially if replies and quote posts are thin relative to the view count, the reach was likely lifted by distribution. As a reference point, the 10 organic launches in this corpus carry about 486 replies and quote posts per million views against about 230 for the amplified set. FORKOFF runs this same read on every view of a launch it delivers, so the reach it reports is reach an audience actually saw.
The corpus is 30 public launch posts published between October 2025 and June 2026, spread across 6 market verticals, carrying 110.4M+ views in total. A launch enters it because its metrics were public and complete and its product name unambiguous, which is not a random draw from every launch on X. So the 67% figure describes these 30 launches and should not be read as a population estimate for the platform.
Authorship
Simba
Co-founder, FORKOFF
Reviewed by: Kshitij JK
Last reviewed:
Published:
Methodology
The X Launch Authenticity Study 2026: a first-party data study. FORKOFF RADAR scored 30 public X product launches published between October 2025 and June 2026 on the views-per-like method, dividing each launch post's public view count by its public like count and reading the ratio against a calibrated ladder (a roughly 500 organic ceiling, a roughly 2,000 heavy-distribution line, a roughly 5,000 fraud-tier band), with replies and quote posts weighed to keep a light read from being mislabeled heavy. Distribution: 10 organic, 20 carrying a distribution-amplified signature (about 67%). Confidence: 18 verified traces, 12 reconstructed reads, and a reconstructed launch can never carry a high-confidence amplified verdict. Counts are computed from the underlying data on every build. Reported at the aggregate level; the only launches named are the ones that read cleanly organic. Amplified reach is framed as legal and common; RADAR reads a signature, never an intent.
Sources cited
FORKOFF delivers product launch videos to a guaranteed view tier through organic distribution, then audits every view with the same views-per-like method this study runs on the public corpus. See how a guaranteed-view launch works.
Explore the launches
HyperAgent launch
@howietl
14.3M views · 2,265 per likeRead
SubQ launch
@alex_whedon
13.1M views · 568 per likeRead
OpenAI's Jalapeño AI chip launch
@OpenAI
7.1M views · 312 per likeRead
Cursor for iOS launch
@cursor_ai
6.4M views · 492 per likeRead
Anoria launch
@michaelblhssn
5.7M views · 4,410 per likeRead
Meridian launch
@alexmdees
5.3M views · 1,523 per likeRead
The benchmark behind every reading
OpenAI's Sora beat ChatGPT to 1M downloads, then shut down 6 months later. The public download, revenue, and moderation data behind both halves.
Read
On an ordinary post a pod costs you some reach. On a launch it costs you the only measurement you had. LinkedIn's own trust VP explains where the penalty lands.
Read
LinkedIn's own policy forbids agreeing ahead of time to like each other's posts. Here is what earns distribution instead, and why the cadence data disagrees.
Read
