Corrections

How this site corrects errors and answers panels that dispute a result, and a log of every correction.

Rules

A factual error is corrected, with a date, and never silently. Every correction is logged on this page. Silent editing is what the rest of the niche does and why it is not trusted.
Disagreement with the method does not change a result. The method was published before measurement; dispute it publicly.
A removal request is refused, but a right of reply is offered. The panel's statement is published beside the finding, unedited and uncommented.
Response time: five working days.
The archived snapshot is the defence. A panel claiming it never published something gets a link to the snapshot from the day of measurement.

The author decides; both the request and the outcome are logged below.

To report an error, write to the address on the contact page.

Log

Date	Page	What was wrong	Outcome
4 November 2026	/what-happens-to-your-balance-when-a-panel-goes-quiet/	The guide said, in the present tense, that the panel this site is operated by publishes nothing on whether its balance expires or what happens to it when an account is closed. That was true of the page read on 1 October 2026 and is not true of the page now: a section was added to the refund policy after that reading, and the copy archived on 4 November 2026 states both.	The sentence is in the past tense and dated to the 1 October reading, which is the day the audit measured, and a paragraph after it says what the page states now and links the copy archived on 4 November 2026. The October result is unchanged: it is a reading of the October page, and the January audit reads what is published then. The change list on the audit gains a row for the 4 November rewording, which changed the words and not the commitment.
2 November 2026	/faq/	The answer to "Does this site measure what panels actually deliver?" opened with "Not yet" and said that until the first results are published, no ranking here uses delivery data. The first purchase test was published on 1 November 2026, so the answer was wrong from that day.	The answer now opens "Yes, by buying", gives the day Test 1 was published and what it measured, and says that the first delivery score, read from Test 1 and Test 2, is published on 1 December 2026. The clause about no ranking using delivery data is kept, because that is still true and now has a date on it.
2 November 2026	/methodology/	The page said the delivery score "is published monthly and grows as each new service is tested". No delivery score existed: no formula had been fixed, no criteria file had been written for one, and nothing on this site published a score monthly. Two further sentences, on Test 1 and on Test 3, still spoke of the first delivery test as something to come after it had been published.	The delivery score is published with the tests and not on a calendar, the first one on 1 December 2026 from Test 1 and Test 2. Its four criteria and their weights were fixed by Zane Roberts on 2 November 2026, after Test 1 was public and before he had seen Test 2's figures, which Jack counted on 22 October and hands over on 23 November; Jack does not review or change the formula. The page now prints the criteria in a table rendered from the locked file the score reads, so the published formula and the score cannot move apart. The half that criterion carries is a round number chosen after Test 1 was public, and the page says so and says that on Test 1 it is passed by MoreThanPanel alone and failed by Boostero. The two sentences about Test 1 are in the past tense and link to it.
2 November 2026	/about/	The page said the purchase test of what panels deliver "publishes on 1 November 2026". It was published on that day.	The sentence is in the past tense and the words that carry the date link to the test.
2 November 2026	/audit-2026-10/	In "What we did not measure", the first purchase test was described as something that "is published on 1 November 2026, and until then no ranking on this site uses delivery data". It was published on that day, and the audit left a reader no way to reach it.	The sentence is in the past tense, the date links to the test, and it now names the day the first delivery score is published, 1 December 2026, which is what has to happen before any ranking here uses delivery data. No figure or result in the audit changed.
2 November 2026	/methodology/	The page said twice that a reader could check it in a public commit history: that both versions of the October 2026 criteria are there, and that every version of the page is there. Neither is reachable. The criteria files and the page itself live in a private repository, and the public repository holds the archived panel pages and the measurement files, which name a criteria file and a commit but do not contain either.	Both sentences are gone. The criteria section now says that the table is the version applied to every panel, that the version it replaced is not published, and that this page is the record of the revision and its date. The changes section now says that every change to this page is dated in the table below, with the kind of change named. Nothing is claimed about a commit history a reader cannot open.
19 October 2026	/does-an-smm-panel-check-the-link-you-paste/	The finding said the API document the ten panels share describes the same validation, and its FAQ said the same validation order is what their documents describe. It does not. The document archived on 1 October lists what the add call takes, a service, a link, a quantity, runs and an interval, and the only error it gives as an example is "Incorrect order ID": the words valid, check and insufficient do not appear in it at all.	The body and the FAQ now say that the document lists the parameters and says nothing about which of them is checked or in what order, so it cannot settle the question for any panel. The FAQPage data carries the same words. The claim rested on one panel's API, which is what the finding measured, and that has not changed.
19 October 2026	/snapshots/2026-10-01/	An evidence link to a gzipped archive opened as binary. The file is stored gzipped and was served as text with a gzip encoding, and the edge compresses what it serves, so it arrived encoded twice. The decompressed copy that the audit and the finding link to did not exist.	A gzipped snapshot under eight megabytes decompressed is now published as .html.txt beside it, and the links in the audit and the finding point there. A gzipped file is served as a download that gunzips to exactly the bytes index.json hashes, and the day page marks those links. Two catalogues are above the limit, at 41.7 MB and 12.8 MB decompressed, which is more than the host serves: they stay gzip downloads and their links say so.
15 October 2026	/faq/	The answer to "Does the panel check the link I paste?" said a form may check that the field is a URL, and that a profile link pasted into a likes service tends to fail later at the supplier rather than on the form. No form was ever measured, and the panel record of the order placed on 4 September 2026 shows the opposite of the second half: the order was accepted with the link "notaurl", charged, and later reported completed with nothing returned.	The answer now gives only what was measured: the API checked the service ID, the quantity range and the balance, carried the link without inspecting it, returned an order ID and charged. It publishes with the finding at /does-an-smm-panel-check-the-link-you-paste/, which sets out the one order it rests on.
14 October 2026	/methodology/	The method said the day 21 count of the first delivery test would be taken on 13 October 2026 at 11:00, and that the refill requests of the second test would be filed an hour later at 12:00, all ten inside that hour. Neither time held. The count was taken at 17:40 CEST and the ten requests were filed between 17:49 and 17:57 CEST, nine to seventeen minutes after it.	The page now gives the times that were kept, and says that both ran later in the day than the protocol named. What the protocol required of the requests did hold: one per panel, identical wording, none chased, and all ten inside eight minutes of each other, which is tighter than the hour the rule allowed. The changelog line of 6 October is marked amended where it gave the planned times. The figures are recorded in measurements/2026-10-13/ and are checked against the workbook when it arrives on 23 October 2026.
12 October 2026	/glossary/real/	Four glossary entries claimed more than this site can show. /glossary/real/ said nobody in the chain verifies the label; /glossary/ticket/ said a ticket never reaches whoever delivered the order; /glossary/service-id/ said the same service carries a different number on every panel in the chain; and /glossary/comments/ said nobody read the post and reacted to it, which states what readers did rather than what the service is.	Each now says what can be seen: nothing in the chain requires anyone to verify the label; a ticket does not reach the deliverer directly; the number is usually different on each panel; and the comments are not a reaction to the post, because the text was ordered. No figure and no result moved.
10 October 2026	/glossary/cancel/	The block said a cancelled order does not send money back to your card or wallet, and said it without qualification. Some panels do return funds to the original payment method on request, and this site's own entry at /glossary/refund/ says a refund is normally credited to the panel balance and that terms often restrict or refuse a card refund. The claim was published in that form before today and was rewritten yesterday with the word never, which made it worse.	It now says the money comes back normally as credit on the panel rather than to the card or wallet it came from. Two other entries lost a universal they could not support: /glossary/completed/ said nobody checked the delivery and now says the status is the supplier's report rather than a check of your profile, and /glossary/quality-tier/ said no independent party tests the accounts and now says nothing in the tier shows that anyone outside the panel tested them.
9 October 2026	/what-refill-means-on-an-smm-panel/	The corrections of 8 October gave the refill result of the 1 October audit as six panels stating a window and a trigger and "four stated neither". Two of those four do state a trigger and no window, Just Another Panel and Peakerr; only two mention refill nowhere on the pages read, YoYo Media and SMMFollows. The wrong split went out in the guide's FAQ answer, in the measured paragraph on /glossary/guarantee/, and in the outcome of both 8 October rows below. Two sections of the guide also still described the 10 September reading: "What the audit found" attributed its figures to the 1 October audit, and the limits paragraph carried the old date.	All four places now give the split as six, two and two. The guide's "What the audit found" is rewritten from the 1 October reading, naming the pages it took (services, terms, homepage) and the four panels in the two smaller groups, and its limits paragraph carries the right date. The two rows below are marked as amended rather than rewritten in silence.
9 October 2026	/corrections/	The outcomes of the corrections dated 2 and 5 October sent a reader to "the rankings page" for the list of Boostero's changes and for the reason the operator is not ranked. That content is on /audit-2026-10/, and the rankings page is only the index that links it. The two rows were then edited in place to say so, which is the silent change rule 1 of this page forbids.	Both rows now point at /audit-2026-10/ and carry the words "amended 9 October 2026" where the text changed, and this row says what changed and why. A site path written in a correction is also rendered as a link now, so a row can point at the page that carries the explanation rather than only at the page that was corrected.
9 October 2026	/faq/	The answer to "What does API v2 mean?" said the list of actions still differs between panels. The finding published on 1 October read the API document of all ten panels in the set and they are one document, with the same calls, down to the example response: the answer contradicted this site's own measurement.	The answer now says that on the ten panels this site read the document and the calls were the same on all ten, because they run the same panel software, and keeps the caution for panels outside that set. The question links the finding rather than the API guide.
9 October 2026	/tests/delivery/	The section described the delivery tests as "one service a month". They are not: the second test files refill requests on the orders the first test placed and buys nothing, and the third is a different service ordered on 24 November 2026. The line stating when the first test publishes was also printed twice on the page.	The description says orders counted at fixed checkpoints, without a monthly cadence, and the repeated line is gone. The three tests and their dates are set out on the methodology page.
9 October 2026	/are-top-smm-panel-lists-trustworthy/	The guide said that an audit had recorded ChatGPT describing a list as "promotional rather than independent evidence" and moving it down its answer. The quotation came from the list published on Boostero's blog, and this site holds no dated record of it: it was presented as something recorded when nothing here records it.	The quotation is gone and will not be restored without a date and a source. The paragraph and the limits note now rest on the one reading this site does hold, dated 11 September 2026 on the methodology page. The guide also says the publisher's own panel usually ranks first rather than always, in line with its own body, and no longer reads as though the operators' panel were an eleventh panel beside the ten.
9 October 2026	/faq/	The answer to "What is start count?" said every refill claim is settled by comparing the current count with the start count plus what the panel delivered. The usual baseline is the start count plus the quantity ordered, which is what the glossary and the refill guide say, and not every panel settles a claim that way.	The answer gives the ordered quantity as the baseline and says a claim is usually settled that way rather than always. The refill guide, the glossary, the API guide and the drip-feed guide now state the same baseline. The three points a refill window can start from are on the refill guide and the two glossary entries; the API guide states no window and the drip-feed guide counts from its own runs (amended 9 October 2026).
8 October 2026	/what-refill-means-on-an-smm-panel/	The guide answered "how many panels state their refill terms before you order" with a figure from a list published on Boostero's blog on 10 September 2026, and cited that list as one of its two sources. The set behind that figure included a retail seller this site puts out of scope, so it was not ten panels of the kind this site measures, and the one panel said to state its terms per service was Boostero, which the sentence did not say.	The figure and the link are gone. The answer now uses this site's own audit of 1 October 2026, where six of the ten panels state a refill window and a trigger before payment (amended 9 October 2026, see above: the other four do not divide that way), every reading links the archived page behind it, and the text says that the panel this site operates is among the six.
8 October 2026	/are-top-smm-panel-lists-trustworthy/	A guide about who writes ranking lists cited, as its source for what AI assistants say, a ranking list published on Boostero's blog which puts Boostero first and ends in a sign-up button: the pattern the guide itself warns about.	The link is gone. The point now rests on the reading recorded on the methodology page, dated 11 September 2026, of what ChatGPT gave as its first reason on a reseller prompt.
8 October 2026	/glossary/guarantee/	The measured paragraph gave two counts of the word guarantee, 24 and 17, from a list published on Boostero's blog on 10 September 2026. Two of the four panels counted there were quoted and the other two were not, so the two highest numbers stood for the set.	The paragraph now carries this site's own measurement of 1 October 2026: six of the ten panels state a refill window and a trigger where a buyer reaches them before paying (amended 9 October 2026, see above). The glossary may now cite an audit published here, which carries the archived page behind every figure.
8 October 2026	/about/	The ownership disclosure in the footer of every page, on the About page and in the FAQ said Boostero is "scored by the same published formula as every other panel" and labelled wherever it appears in a ranking. Since 5 October 2026 it is scored and not ranked, and it is labelled on every page of this site, not only in a ranking.	The disclosure now says it is measured on the same criteria, on the same day, as every other panel, that it is labelled wherever it appears on this site, and that since 5 October 2026 it is scored and not ranked.
7 October 2026	/audit-2026-10/	The article's own FAQ answer to "Can a panel change its score?" said no panel is told its result before publication. That is not true of Boostero, whose operator runs the audit, and it contradicted the disclosure above the table on the same page. The FAQ on /faq/ had been corrected for this on 2 October; the copy inside the article had not.	The answer now states the exception, names it as the reason Boostero is scored and not ranked, and points at the disclosure above the table. The FAQPage structured data carries the same words as the page.
7 October 2026	/audit-2026-10/	In "What we did not measure", the sentence "For those five the services criterion is unverified rather than passed" had nothing to refer to: the sentence naming the five panels whose services URL answers with the homepage was cut on 6 October, when the article was shortened to fit its word limit, and this one was left pointing at it.	The sentence now says which five it means, by what they have in common rather than by listing them again.
7 October 2026	/audit-2026-10/	"Legal identity" said three of the ten panels name a legal entity and a jurisdiction. Peakerr names one too: the footer of every page this audit archived carries "DIGITAL SOFTTECH LIMITED", a registered address in Ruislip, London, and "Company No: 14815184". The reading had missed it. The quotation given for bulkfollows also broke off mid-number, at "Company No".	The count is four, Peakerr is listed with the words from its footer and a link to the archived page, and the bulkfollows quotation now carries the whole registration number. The other six panels were re-read against the same measure, Boostero among them, and none of them names an entity. The finding is not scored, so no score moved.
7 October 2026	/audit-2026-10/	MoreThanPanel's key highlight said its terms state that deposits do not expire. That sentence is in its refund policy, which is the page the body of the article cites for it.	The highlight now says refund policy, matching the body and the archived page behind it.
7 October 2026	/audit-2026-10/	Under the results table the page said reviews of single panels are linked from their row and listed on the reviews page. No review has been written, so nothing was linked from any row; the rows link to the panels themselves. The same sentence was on /methodology/.	Both now say that where a review is written it is listed on the reviews page and linked from that panel's row.
7 October 2026	/audit-2026-10/	The disclosure above the table gives the SHA-256 of the catalogue read on 30 September 2026, but that snapshot was only in the public evidence repository: /snapshots/2026-09-30/ answered 404 on this site, so a reader following the hash had nowhere to check it.	The whole 30 September rehearsal run is now served under /snapshots/2026-09-30/, 44 stored pages whose bytes hash to what its index records, and the sentence links the Boostero catalogue it quotes.
7 October 2026	/audit-2026-10/	The article said the set was selected on 16 September 2026 while the figures behind the selection, published on /methodology/, carry 16 and 18 September. A reader could not tell that part of the table was read after the choice was made.	The FAQ answer and the methodology page now say the set was chosen on 16 September and that each figure carries the date it was read, 16 or 18 September, so a reading taken after the choice is visible as one.
7 October 2026	/snapshots/2026-10-01/	The page said "55 pages, fetched between 13:26:12Z and 14:29:34Z". Fifty-three pages were stored and two were not found at any address tried, and the last fetch was at 15:32:33Z: the run wrote its finishing time when its first pass ended, and the pass that filled the gaps ran after it.	Every snapshot day now counts stored and not-found pages separately and reads the window from the pages themselves rather than from the run's own header.
7 October 2026	/audit-2026-10/	The section headed "Compared with our prior audit" called a list published on Boostero's blog "our audit" and "the earlier audit". It is not an audit: it was not made under the criteria, the weights or the archive rules this site measures by, and it ranks the panel this site operates first. The same wording was on /methodology/.	Both now name it for what it is, a comparison list published on Boostero's blog on 10 September 2026 that ranks Boostero first, and say that it is not a measurement and that no result here is read against it.
5 October 2026	/faq/	The answer to "Can a panel pay to move up?" and the answer to "Why should I trust a ranking published by a panel?" both said the panel this site operates is scored by the same formula on the same day, which was true and left it standing first on its own list.	Both answers now state that it is scored and not ranked. The rule changed on 5 October 2026, after the October measurement and with the scores known; /audit-2026-10/ explains the change above its table (amended 9 October 2026, see above) and publishes the score it would have ranked on.
2 October 2026	/faq/	The answer to "Can a panel pay to move up?" said no panel is told its result before publication. That is not true of Boostero, whose operator runs this audit, sees every reading, and changed its catalogue after the rehearsal read of 30 September.	The answer now states the exception and points at the list of Boostero's changes at the top of /audit-2026-10/ (amended 9 October 2026, see above). Published with the October rankings, so the FAQ never points at a list that is not there yet.
1 October 2026	/faq/	The FAQ answer to "Do SMM panels have webhooks?" said most panels do not, and promised a finding for 1 October 2026. The finding published that day read all ten API documents and none of them offers one, so the answer understated what had been measured.	The answer now states that none of the ten does, gives the reason the documents are identical, and links the finding.
1 October 2026	/do-any-smm-panels-support-webhooks/	The finding said several of the ten documents accept a list of order IDs in one status request, while its own argument is that the ten publish the same document.	Counted again in the ten archived documents: all ten describe a multi-order status call. The article now says all ten.