Why We Withdrew Six Quote Entries
The October 10 quote withdrawals explain the gap between a recognizable scene and evidence for a particular English sentence.
On October 10, 2026, we withdrew six quotation entries whose exact English wording lacked an adequate witness. That decision means we could not support publishing those sentences as checked quotations. It does not establish that the sentences were invented.
Recognizing a scene did not settle the wording
The review concerned entries associated with Gundam, JoJo's Bizarre Adventure and Berserk. Their recorded citations led to uncited transcriptions without an identified English rendering. The audit also considered related scene descriptions, circulated variants and fan-captioned material. None supplied the qualifying replacement needed for the exact words being published.
A description can establish the subject of a conversation without preserving its dialogue. A different English rendering can make the same general point without verifying the sentence on a quote card. And a search result that repeats this site's own text cannot independently justify the text it copied.
| Evidence encountered | Question left open |
|---|---|
| A recognizable scene | Does this identified English release use these words? |
| A circulating variant | Which witness actually prints the version being cited? |
| A caption attached by a fan | Where is the corresponding licensed or qualifying editorial source? |
| No qualifying result | Could an unread release still establish the wording? |
Withdrawal reached beyond the quote list
We removed the six entries and their twelve quote-card images. The review also removed dependent passages from the JoJo roundup and Berserk ending guide, and unsupported dialogue material from Amuro's profile. The character addresses remain available; Amuro's page explains that verified quotations are unavailable there.
That wider check matters because an unsupported line can survive after its main entry disappears. A roundup might repeat it without quotation marks, or a character introduction might build an interpretation on it. Removing the original entry is not enough if the same unsupported assertion remains part of the site's argument elsewhere.
What could change the decision?
Restoration needs the exact wording in an identified licensed English release or a qualifying editorial body quotation, with attribution and a usable locator. The surrounding commentary would then need review against that evidence. Finding a source is an opportunity to reconsider the entry, not a reason to restore every old claim around it automatically.
We retain the previous records so that a later check can compare like with like. The public fact-check index distinguishes verdicts on the claims it covers; a withdrawn quotation should not be silently converted into a proven fabrication. The verification policy describes that distinction.
These six decisions illustrate a limited but necessary conclusion: a site can lack enough evidence to publish a quotation while also lacking enough evidence to declare it false. Withdrawal respects both limits.
When the trail comes back to itself
Try this invented loop: page A cites page B as its authority, while B’s source link returns to A. Following both links reaches no observation outside the pair. Add a third page quoting B and the loop still supplies no episode or printed page. Record the circular path as A → B → A, then look for a witness outside it. This is a practice example, not an allegation about named sites.
Recording a search without a match
Here is an invented search-log example, not another withdrawal from this site: “Candidate wording: [short excerpt]. Checked: [named edition and pages]. Result: no match within those pages. Related result: [a scene summary], which describes the event but supplies no dialogue. Still unread: [another identified release]. Next useful evidence: the passage in that release.” The brackets are fields to fill, not sources already checked.
When assessing a quote list, make the same distinctions for each entry. Does its link reach body text containing the words? Does that body identify the speaker? Is the English rendering named? Does the claim extend beyond what the source supplies? Does another apparent witness merely repeat the first? Record the answer rather than awarding confidence for the number of links. An inaccessible page belongs in the log as inaccessible; it is neither a confirmed match nor a demonstrated absence.
A source trail can end without an answer
Consider an invented citation chain: a list copies a fan caption, and nine other pages copy the list. The ten pages are ten places to encounter the claim, but their agreement supplies no additional observation of the scene. Recording who cites whom makes that limit visible.
Two dates belong in such a log: the earliest appearance actually found and the date of the check. Neither should be renamed the quotation's origin without further evidence. An older unread release can remain outside the search.
Keep access status separate from the wording judgment. A saved page inspected earlier and an unsuccessful live recheck are different observations; neither should silently replace the other. A dead link is an access problem, not positive proof that the quotation was invented. Record what the retained copy supports and what can no longer be confirmed live.
When a result contains only part of the sentence, leave the missing words unresolved. A familiar variant cannot complete the witness. The same rule applies to an editorial summary: adding quotation marks does not convert a writer's paraphrase into a character's exact speech.
Reading an archived copy of a dead page
An archived capture can preserve what a page once said, and its address records when. The Internet Archive's Wayback Machine help explains that the number in the middle of an archived URL is a date code, yyyymmddhhmmss, and that archived pages can be linked and cited. It also warns that an archived site may be incomplete: for a missing link the tool takes the closest available date, or the live page if none was archived. A capture can therefore mix dates, so record the date code of the page that actually holds the cited words. The help page adds that a blue calendar dot marks a capture whose server returned a good response, while orange and red mark errors. A capture of an error page is not the source; it is evidence that the source was unavailable when crawled.