A 99-minute comparison during a large YouTube LIVE

On November 14, 2026, we tested the production version of TikChat alongside the native chat of one large public YouTube LIVE. Across a 99-minute observation window, we collected 37,104 distinct text messages in the native browser sample and matched 37,100 to the real TikChat operator view. Four entries remained unresolved.

This was a broadcast with a substantial existing audience. A pre-test reading of the YouTube watch page showed approximately 85,000 concurrent viewers. That single reading establishes the scale of the broadcast at that point; it is not a measured peak or an audience average. Its scope is documented in the audience context note.

The ID result corresponds to 99.989% agreement within the observed sample. We also compared message content separately, so finding the same ID was not treated as proof that every word and emote representation agreed.

This was our own product test, commissioned by Matt Olech, the owner of TikChat. The owner started the production session; an AI assistant observed both browser views and performed the comparison. The broadcast, channel and viewer identities are omitted from this report.

Test setup

We connected one existing public YouTube broadcast to a production TikChat session. The reference was YouTube's native browser chat, set to Live chat. The application view was the already-open TikChat operator, with YouTube as the source under test.

  • Study date: November 14, 2026.
  • Platform: YouTube only.
  • Broadcast context: approximately 85,000 concurrent viewers in one pre-test browser reading.
  • Application: the real production TikChat operator.
  • Reference window: 99 minutes and 0.299 seconds.
  • Final comparison period: a further 97.248 seconds, reading TikChat only.
  • Unit: a distinct native YouTube text-message ID observed after excluding the starting buffer.
  • Selection: one available public broadcast, without random selection or a comparison group.
  • Published evidence: aggregate results, relative sampling times, methodology and measurement code, without comments or viewer identifiers.

The observation window is not the complete duration of the YouTube broadcast. Sampling began after the broadcast and the TikChat session were already running. The owner confirmed the broadcast's natural end; its exact start and end were not independently timestamped.

What was the purpose of the test?

The question was whether text messages observed in a busy native YouTube chat could also be found in the production TikChat view, and whether their displayed content agreed under a defined comparison method.

That required two separate checks: message identity and message content. An ID match establishes that the same reference message was found in TikChat. A content comparison asks whether the browser representations of that message agree, including the text labels used for inline emote images.

The broadcast's large audience made this a useful real-world setting. The measured outcomes were the sampled messages and their matches, rather than the number of viewers TikChat could serve or any change in audience engagement.

Method

Establish the native chat and production session

The owner started the YouTube source in TikChat. The observer reused that operator tab and opened the native YouTube chat separately. The native view was set to Live chat, and the browser structure and ID transformation were checked before formal collection.

A calibration found 205 shared message IDs. Of those, 203 also had matching normalized content, while two involved differences with emotes. These calibration entries were not reported as additional study results.

Exclude the initial reference buffer

A newly opened chat can contain messages posted before observation begins. We recorded the starting buffers and excluded 212 initial native YouTube IDs from the new reference count. The initial TikChat buffer could still confirm an entry that became visible in the native reference later.

Read both browser views repeatedly

The observer aimed to read both chats about every five seconds. The completed record contains 932 native reference reads and 945 TikChat reads after the baseline. The latter includes 13 reads during the final comparison period.

Each reference read inspected up to 500 recent native chat rows. Each TikChat read inspected up to 800 recent YouTube message rows. Reading the same ID again did not increase the count of distinct reference messages. Actual read intervals, buffer overlap and exclusions were logged.

Match IDs and compare message content

Matching used provider message IDs transformed into the form represented by TikChat. It did not rely on similar wording, usernames or AI judgments about meaning.

For content, the checker read the message text and the alt labels of inline emote images. It normalized Unicode to NFC, collapsed consecutive whitespace and trimmed whitespace at the edges. Letter case and punctuation remained significant. Temporary hashes allowed comparison without retaining the comment text in the evidence files.

The checker distinguished matching normalized content, differences involving emote representation with matching ordinary text, and other unresolved content differences. It also checked for differences limited to emoji variation selectors; none were classified in that category.

Save aggregates and complete the final comparison

We saved progress aggregates approximately every five minutes. An unmatched reference entry was initially pending; after 60 seconds without a match, it was classified as unresolved. It could still be matched later.

After the owner confirmed the natural end, we stopped adding reference entries and continued reading TikChat for 97.248 seconds. The final ID totals did not change during that period. Working message pseudonyms were then cleared, leaving the aggregate evidence and sampling record.

The measurement used local browser reads. The observer did not start another TikChat collector, add a separate TikChat API polling loop, connect to the VPS, deploy code, restart services or change application settings. Server resource use was not measured.

What happened during the observation?

The sample grew from 1,731 entries at the first five-minute progress snapshot to 37,104 at the end. Entries that were still pending at a snapshot were kept separate from those that had remained unmatched beyond the observation threshold.

The first progress snapshot with an unresolved entry was the one labelled 50 minutes. The 65-minute snapshot showed two, and the 70-minute snapshot showed three. The final comparison retained four unresolved IDs. These are sampled status observations, not exact timestamps of when a reception problem began.

The native YouTube view remained the reference. Sampling pauses and buffer replacement were retained in the log. The observer also checked the TikChat view for active search filtering; accepted reads had an empty search field. A separate control read found the Hide A-D replies filter switched off.

The working comparison recorded zero browser-read errors. That describes successful reads of the pages; it does not prove an uninterrupted provider connection or complete reference coverage.

Results and evidence

Did sampled YouTube messages appear in TikChat?

37,100 of the 37,104 distinct native reference messages were matched to the production TikChat view. Four remained unresolved. The resulting 99.989% is agreement within this browser sample, not the percentage of all comments captured from YouTube.

Sampled native YouTube text-message IDs compared with the production TikChat view

ID comparisonEntries
Distinct sampled native text messages37,104
Matched to the production TikChat view37,100
Unresolved at the final comparison4
Still pending at the final comparison0


All reference entries included in these totals were classified as ordinary text messages. This sample does not provide separate results for Super Chat, paid stickers or membership activity.

How did the recorded result develop?

Cumulative results at selected progress snapshots and final comparison

Progress snapshotReference sampleMatched at that pointPendingUnresolved
About 5 minutes1,7311,674570
About 25 minutes9,7409,691490
About 50 minutes18,90218,878231
About 75 minutes28,22128,200183
About 95 minutes36,98336,968123
Final comparison37,10437,10004


The intermediate rows are progress snapshots taken at approximately the labelled times. Their pending and unresolved counts can change as later matches are found. The downloadable record contains the actual snapshot times and all five-minute updates.

What does an unresolved entry mean?

An unresolved entry is a native reference ID for which this method did not establish a TikChat-view match by finalization. It is not automatically a message proven lost by the backend. Reception, presentation, moderation, buffer behaviour and observation limits remain possible areas for investigation. This test did not assign a cause to the four cases.

Did the message content agree?

Among the 37,100 ID-matched entries, 36,346 had identical normalized rendered content, or 97.968% of the matched set. Another 622 had matching ordinary text but differences involving emote representation. The remaining 132 pairs had other unresolved content differences.

Content agreement among 37,100 matched message IDs

Content comparison among matched IDsEntriesInterpretation
Matching normalized rendered content36,346Text including inline-image alt labels agreed under the defined normalization
Matching ordinary text, different emote representation622Image-label or representation differences; visual equivalence was not established
Other unresolved content differences132The comparison did not establish agreement or a specific cause
Total ID-matched entries37,100The denominator for the content comparison


These categories sum to 37,100. The four unresolved IDs have no matched pair for a content check. If all 37,104 reference entries are used as the denominator instead, normalized content agreement is 97.957%.

The distinction matters for emotes. Two interfaces can use different text labels for an image. A difference in those labels does not by itself establish a different picture or a corrupted comment. Equally, the 622 cases cannot simply be added to the fully matching category without further verification.

A diagnostic spot check of 191 currently shared entries found six content differences: five with matching ordinary text but different image labels, and one other case. This was a check of the current buffer, not a reclassification of all differences in the study. The 132 other differences remain unresolved.

Evidence files and verification

The evidence files document the counts, sampling process and comparison rules for this test. They omit the broadcast title, channel link, viewer names, comment text, raw message IDs and account identifiers. Times are relative to the observation start.

The local verification tools check arithmetic, timing consistency and file integrity. The instrument's 16 tests use synthetic data; passing them is not independent validation of the real broadcast.

The aggregates can be checked and the code can be used for a fresh observation. The original message-by-message matching cannot be independently replayed after disposal of the working data. A checksum detects a changed file against the supplied checksum; it is not third-party certification of the measurement.

What does the evidence support?

This test supports a specific observation: during one large public YouTube LIVE, 37,100 sampled native text-message IDs were also found in the real production TikChat operator. It also provides a separate account of content agreement, emote-related differences and unresolved cases.

The scale comes from two distinct observations: a large existing audience in the pre-test reading, and a reference sample exceeding 37,000 messages. Those facts make this a substantial single-session product test. They do not turn the audience into 85,000 TikChat clients or establish a general capture rate for other broadcasts.

Limitations

The native reference was sampled

The median interval between native reads was 5.086 seconds, the 95th percentile was 8.103 seconds, and the longest interval was 75.053 seconds. Eighteen successive nonempty native buffers had no shared IDs, signalling possible observation gaps. The number of messages missed by the observer is unknown.

The first formal reference read began 51.610 seconds after the observation start. The last ended 48.967 seconds before the reference window closed. The 99-minute window therefore must not be described as continuous recording of every YouTube message.

Both browser views had bounded buffers

The TikChat reader deliberately used only the latest 800 matching rows. The log marks that bound as truncation; it is not a count of application failures. Unsupported native DOM rows were excluded from the message denominator. Their repeated appearances in the log are not distinct comments or proven lost messages.

Some content representations changed

The observer recorded 259 changes of the reference content digest for already-known IDs, and zero such changes in the operator view. This counts changes, not necessarily 259 different messages. The checker kept the latest observed content representation. The causes of those changes were not established.

The final content result is consequently a comparison of observed browser representations under the stated rules. It is not a manual visual inspection of every emote or a finding that all 132 other differences were defects in TikChat.

Observation timing was not transport latency

Fifty-two IDs were first found in TikChat more than 60 seconds after their first native observation. Read intervals and pauses by the observer contribute to that difference. It cannot be used as a count of messages proven delayed by the transport, or as a precise latency benchmark.

This was one first-party production session

The owner commissioned the test, and an AI assistant performed the observation. There was no independent reviewer, random sample of broadcasts or control group. The audience reading was a single pre-test context point, not a peak measurement or an audience time series.

The evidence is from the browser views. We did not independently audit the production database, record a production release fingerprint or measure server CPU, RAM, I/O or network use. It is therefore not a server load benchmark. The detailed messages and working pseudonyms are not retained for later message-level re-auditing.

What this means for creators

For a creator considering TikChat for YouTube, this test offers a concrete account of the production tool during a large, existing LIVE. It documents more than 37,000 observed reference messages, the IDs found in the operator and the separate outcome of comparing content.

The practical evidence is that the real TikChat view presented the same identifiable conversation entries throughout the sampled observation, with four unresolved reference cases. The content table makes the remaining work visible instead of hiding emote representation and other differences behind a single success percentage.

The useful conclusion is specific to this session. It provides evidence for evaluating the YouTube chat view, without claiming that TikChat changed the broadcast's audience, engagement or creator response rate.

Frequently asked questions

Was this a test of the real TikChat application?

Yes. The owner started a real production TikChat session with YouTube. The observer read that existing operator tab alongside the native YouTube chat. The application was not replaced with a local copy for the measurement.

How large was the YouTube LIVE?

A single pre-test reading of the native watch page showed approximately 85,000 concurrent viewers. It was a large public broadcast, but the study did not record a peak or average audience. Its formal comparison concerned 37,104 sampled chat messages.

What does 99.989% agreement mean?

It means 37,100 of the 37,104 distinct text-message IDs observed in the native sample were found in TikChat. Four remained unresolved. Because the native reference had sampling gaps, this is not the percentage of all messages captured from the broadcast.

Did the comments themselves agree?

For 36,346 of the 37,100 matched IDs, normalized rendered content agreed. Another 622 had matching ordinary text with differences involving emote representation; 132 had other unresolved content differences. Identity agreement and content agreement are separate results.

Were four comments definitely lost by TikChat?

The test did not establish that. Four observed reference IDs had no confirmed operator-view match under this method. Their causes were unresolved, and the browser comparison was not an independent audit of the production database.

Was this a load test with 85,000 TikChat users?

No. Approximately 85,000 described the YouTube audience in one pre-test reading. We tested a single TikChat session with a YouTube source. We did not simulate that many TikChat clients or measure the server's load capacity.

Can readers verify the result?

Readers can inspect the aggregate records, sampling log, method, code and checksums, and rerun the arithmetic checks. They cannot independently replay each original message match, because the working message-level data was discarded.

Conclusion

In this first-party test during a large YouTube LIVE, 37,100 of 37,104 sampled native text-message IDs were matched to the production TikChat view over a 99-minute observation window. Four cases remained unresolved. Content was assessed separately, with 36,346 matching normalized pairs and the remaining differences reported explicitly.

The report documents one substantial real-session comparison, including its sampling gaps and limits. It provides a specific, inspectable result for TikChat's YouTube chat view rather than a claim about all messages or every future LIVE.

To connect a YouTube broadcast to the production operator, see the TikChat setup guide.