Introduction
A clear recording can preserve an uncertain account
An interviewee says they remember a cinema closing in late August and might be a week out. A summary says the cinema closed in late August. The change is small in length and large in certainty. The first sentence describes a recollection; the second presents a date as an event fact. Keep what was said distinct from what has been established. A recording can support an accurate account of the source's words without confirming the event they describe. That does not make the interview useless. It identifies the kind of evidence it supplies and the follow-up needed. A dependable summary lets the reader understand the account at that level instead of silently upgrading it. The result can remain direct: the source recalled a late-August closure but qualified the timing. That sentence is both shorter than the exchange and more faithful than a confident statement whose certainty came from removing the speaker's own limitation.
The same problem appears when different kinds of information are compressed together. A source may have seen a barrier, heard about a notice from a neighbor and been asked about roof damage without knowing the cause. A smooth summary can join those pieces into a causal story nobody supplied. Preserve the origin of each claim before writing the narrative connection. AP's published guidance emphasizes attribution for disputable material and accurate, contextual quotations. This guide uses those narrow editorial principles, not a claim of AP approval or a universal newsroom procedure. Its worked case is entirely invented. The useful review question is concrete: does the summary retain the distinction between an observed detail, a recollection, a secondhand report and a possibility introduced by the reporter? If those roles disappear, the note can become more decisive while carrying less information about why its reader should or should not rely on a particular statement.
References: Attribution and quotation context, Preserving source wording
Choose an interview account rather than a finished news finding
Define the deliverable, the source scope and the verification status at the outset. A summary for a reporter's working notes can describe the interviewee's account and identify follow-up. A publishable report needs the reporting and editorial checks appropriate to its claims and newsroom. These are different stages, even when a paragraph from the notes later contributes to the story. Keep that boundary visible rather than make the internal summary sound like a completed article by removing attribution and unknowns. A useful interview note can be concise while saying the source recalled, reported seeing or said they heard something. Those phrases are evidence labels when they reflect the actual source, not unnecessary verbal clutter. They prevent the next reader from assuming that the writer inspected a notice, confirmed a date or established a cause simply because the summary's sentence structure presents those details as ordinary facts.
Use a claim-by-claim account before writing the short paragraph. Identify the relevant source turn, what it asserts, how the person says they know and which qualification changes its meaning. Keep reporter questions in their own role. Then draft the account and compare each material sentence with that working evidence. This is an original editorial workflow, not a validated automatic fact-checking method. A source reference can be present while the attached claim is still too strong. A quote can match exactly while a paraphrase elsewhere removes uncertainty. Review the relationship between the source and the sentence, not just the existence of a citation or timestamp. The working note should make the next reporting step easier to identify. It should not conceal the need for that step by sounding as though all uncertainty was resolved before the interview ended or before the transcript was condensed.
References: AP accuracy and sourcing principles, Separate annotations and source targets
1. Keep the source's knowledge boundaries visible
1.1 Retain a small reporting case before summarizing it
Preserve turn IDs, speaker roles, the question asked, the source's answer, the basis of knowledge, the qualification and the unanswered point. These working categories help a writer avoid combining material that serves different evidentiary roles. The example here concerns an unnamed fictional cinema and was authored for this guide. It contains ten text turns, with five assigned to a reporter and five to an interviewee. There is no real person, place, closure, recording or published story behind it. The case is not a demonstration of a transcription or reporting service. Its purpose is to make editorial transformations inspectable. A real interview requires review of the actual source and any applicable reporting context. If a word, identity or attribution remains uncertain, preserve the uncertainty rather than make the source appear more knowledgeable or confident because the summary needs to fit into a clean paragraph.
The authored interview begins with R01 asking when the cinema closed. W02 recalls late August and adds, “I might be a week out.” R03 asks whether the interviewee saw the closure notice; W04 says no and explains that a neighbor supplied the information. R05 asks whether roof damage caused the closure. W06 says they do not know and only saw the barrier outside. R07 asks how long the barrier remained. W08 reports seeing it on two mornings and says they cannot describe what happened between them. R09 asks about another site. W10 says they heard another site would open but have not checked. These contributions support an account of the interviewee's recollection and reports. They do not establish the closure date, cause, continuous barrier duration or a confirmed opening plan. The summary needs to preserve those limits rather than turn the question sequence into a complete verified story.
References: Source attribution and contextual reporting, Preserving conversational wording
1.2 Separate recollection from a verified calendar date
For the timing claim, retain the recalled period, the speaker's uncertainty, the source turn, the absence of corroboration and the follow-up question. W02 remembers late August but explicitly allows a possible error of about a week. The short account should not give that period the status of a checked closure date. It can preserve the recollection and its precision: the interviewee recalled a late-August closure and said the timing might be a week out. A calendar check could establish whether a stated weekday matches a date, but it would not establish when this fictional cinema closed. In real reporting, a notice, record or other appropriate source might help confirm the date. Identify that later evidence separately. Do not treat the fact that a transcription tool printed a clear date as proof that the event occurred then, or let a more precise formatting choice create an apparently more precise recollection than the source actually supplied.
Approximation can remain useful. It may help locate records, frame another interview question or identify the period needing review. The summary does not need to discard the claim because it is uncertain. It needs to describe the uncertainty faithfully and avoid presenting a narrowed date range as if it had been confirmed independently. If a later check establishes a different date, preserve the difference between the earlier recollection and the verified record in the working account. Do not silently rewrite the interviewee's words to match the record and then quote the altered transcript. Transcription review and reporting verification can improve the same project while remaining different actions. Keeping them separate makes the note easier to trust: the next reader can see what the source remembered, what was later established and where the writer's interpretation or presentation decision entered the account.
References: Accuracy, sources and corrections, Source wording and uncertainty
1.3 Trace secondhand information rather than flattening the source chain
Distinguish the interviewee, the person they cite, what was reported, what was personally seen and what remains unchecked. W04 says a neighbor told them about the closure notice and that they did not see it themselves. A summary that says the interviewee saw the notice changes the source chain. A summary that treats the notice's content as independently established adds another unsupported step. Preserve the chain at a useful level: the interviewee said they learned about the notice from a neighbor and had not seen it. That account can guide follow-up without pretending the follow-up has already happened. The neighbor's identity, exact words and access to the notice are not supplied in the invented case. Do not fill them in to make the account feel complete. A source summary should reveal which link in the chain needs checking rather than quietly remove the chain and present its endpoint as a directly observed fact.
The future-opening claim has a similar boundary. W10 says they heard another site would open but have not checked. A note can record that unverified report while identifying it as a follow-up lead. It should not say a new site is confirmed, and it should not attribute the claim directly to an organizer the interviewee never identified. If additional reporting finds a source in a position to know, record that evidence separately. The original interview remains a source for what this person said they heard. AP's attribution guidance distinguishes information and its origin; the practical workflow here applies that distinction to the whole account rather than only to selected quotations. Keeping the source chain visible helps a writer decide which claims are ready for a public account, which need another source and which should remain in working notes until their basis can be established more clearly.
References: Attribution and original source identification, Sourcing principles
2. Prepare a transcript that does not overstate knowledge
2.1 Review the wording that controls certainty
Retain the recording reference, the untouched draft, the reviewed text, speaker attribution and consequential qualifiers before condensing the interview. Words such as remember, might, heard, not and only can determine the kind of evidence a sentence supplies. If the draft is unclear about them, return to the recording where available. A reporter's question can also be merged into an interviewee's line, making a suggested cause appear to be part of the answer. Review those boundaries directly instead of infer them from the story the writer expects to find. An uncertainty marker in the working transcript is better than a confident statement whose apparent support depends on an unresolved word. The short account can remain useful by narrowing the claim or identifying the specific source check needed. It does not need to resolve every ambiguity through a stylistic choice before another reporter or editor can continue the work.
WhisperAITranscribe, an audio and video transcription tool, can help prepare the draft for this review stage. A news interview summary still requires checking wording, attribution and the basis of material claims. This guide does not establish independent event verification, automatic source corroboration or newsroom approval in the product. The fictional cinema case is a separate text exercise. Keep the distinction between reviewing the transcript and verifying the account it contains. The recording may show that a person clearly said another site would open, while additional reporting is needed to establish whether that plan exists. A clean draft helps locate the relevant passage; it does not remove that reporting question. Carry the source identity and reviewed status when the text moves into another editor so a later summary does not acquire an unsupported claim of verification merely because its wording is precise and its reference looks complete.
References: Accurate and contextual quotations, Documented transcription conventions
2.2 Keep reporter hypotheses out of the source's answer
Distinguish the interview question, the proposed explanation, the actual reply, the stated knowledge limit and the follow-up need. R05 asks whether roof damage caused the closure. W06 answers that they do not know and only saw the barrier. The question introduces a possible cause; the answer does not confirm it. A summary that says roof damage closed the cinema has converted the reporter's hypothesis into an event claim. A summary that says the interviewee blamed roof damage also invents attribution. The supported account is narrower: the reporter asked about roof damage, but the interviewee said they did not know the cause. Keep that distinction even if the possible explanation is the most newsworthy phrase in the exchange. A compelling topic does not become a supported answer merely because it appears in a clear question and a later sentence mentions a visible detail that could fit the proposed story.
The source may still provide a useful observation. Seeing the barrier is relevant to what the interviewee says they witnessed, but it does not identify why the barrier was there or confirm every implication of closure. Describe the observation at its scope. If another source later establishes the cause, connect that reporting to its own evidence rather than retroactively give the interviewee knowledge they expressly denied. The working summary can preserve both the hypothesis and the denial if the reporter needs a follow-up record. A public-facing account may omit an unsupported hypothesis entirely, depending on the reporting and editorial judgment. Those are presentation decisions after the source boundary has been understood. The summary stage should make that judgment easier by keeping the question, answer and knowledge limit distinguishable, instead of compressing them into one causal sentence that leaves no visible sign of how much was assumed.
References: Attribution for disputable information, Accurate sourcing and avoiding distortion
2.3 Use a claim ledger as a working aid, not a truth certificate
A working record can identify the proposed summary claim, its source turns, the knowledge type and the unresolved check. The authored case uses five manually assigned claim records: approximate closure recollection, secondhand notice information, unconfirmed cause, limited barrier observation and an unchecked future-opening report. Each points to the relevant question and answer. These labels are editorial interpretations grounded in a known written case. They are not automated classifications or a rating of a real person's credibility. Keep the ledger small enough to support the actual task. It should help the writer notice where a claim would become stronger than its evidence, not give the impression that filling every field establishes a verified news finding. If a field is unknown, retain that status. A structured note is most useful when it exposes the missing evidence rather than encourages the writer to invent a plausible completion for every empty slot.
The W3C Web Annotation model distinguishes an annotation from the source target it is about. This guide uses only that general separation: a claim note can refer to an interview passage without becoming part of the speaker's words or proof that the underlying event is true. No annotation protocol or fact-checking system was implemented. In an ordinary reporting workflow, a stable passage reference may be enough. The reader should be able to tell what the interviewee said, what the editor classified and what later reporting established. If the interpretation changes, revise the note rather than rewrite the original speech to fit it. Keeping these roles separate also helps collaboration. Another editor can disagree with a claim's strength, inspect the source and change the summary without losing the record of how the earlier interpretation was reached or confusing that earlier interpretation with something the interviewee actually asserted.
References: Annotation and source as separate resources, Source attribution and contextual quotations
3. Preserve the limits of what was observed
3.1 Two observations do not establish a continuous duration
Separate the observed points, the interval between them and the inference about continuity. W08 says the interviewee saw the barrier on two mornings and cannot say what happened between them. That does not establish that it stood continuously for two days, nor does it establish that the mornings were consecutive. A summary can report two observations without supplying a duration. If the duration matters, identify the evidence needed to establish it. The source's explicit limit is useful information, not clutter to remove. The same reasoning applies when somebody saw a vehicle twice, checked a webpage at two moments or noticed a closed entrance on separate visits. The observations can support what was seen at those points. They do not, by themselves, reveal every state in between, and a summary should not create that missing continuity merely because a continuous narrative reads more smoothly.
A separate authored logical control makes the limitation concrete. Imagine five abstract observation slots, with the first and last marked present. Three possible histories fit those endpoints: presence in every slot; presence only in the first and last; or presence in all except the middle slot. Their total marked slots are five, two and four, and only the first is continuous across the full model. The endpoints are identical. This is a checked logical illustration, not a timeline of the cinema or a reconstruction of the interviewee's two mornings. The slots are not actual days, dates or measured times. Its narrow result is that the same limited observations can fit different intervening histories. That is enough to show why a summary should preserve the reported observation scope while leaving duration or continuity unresolved until evidence capable of establishing those properties is available and identified in the reporting record.
References: Accurate attribution of the information available, Preserving source qualifications
3.2 Keep an observation distinct from the explanation around it
A source can report what they saw without establishing why it happened. W06 mentions a barrier outside. The reporter's earlier roof-damage question supplies a possible explanation, but the reply explicitly withholds knowledge of the cause. Keep those roles separate when grouping claims into themes. A paragraph about closure can contain a timing recollection, a secondhand notice report and a barrier observation while labeling each correctly. It becomes misleading when the topic heading makes them seem like corroboration for one completed causal account. The relationship among the details is a reporting question. A summary should not settle it through sentence order alone. If a source says only that they saw something, retain that limit rather than attach a motive, cause or institutional decision inferred from another speaker's question or from the writer's background expectations about what such a scene normally means.
Firsthand wording also remains attributed evidence. The written case assigns the observation by authorship, but a real recording would establish that a person reported seeing the barrier, not automatically prove every aspect of the observation independently. A journalist may have other evidence that supports it; identify that evidence where relevant. The distinction lets the interview remain useful without claiming more than it supplies. It also avoids an unhelpful binary in which a source's account is either treated as established fact or discarded entirely. The summary can retain a meaningful observation, its source and its limits. That gives the reporter a better basis for deciding what to corroborate, which detail belongs in a story and what wording fairly communicates the evidence. The short account should clarify those decisions rather than make them invisible by presenting a collection of differently supported details in one unqualified factual voice.
References: AP accuracy and source principles, Attribution and reporting context
3.3 Keep a future lead separate from a confirmed plan
Distinguish a secondhand report, a named source in a position to know and independent confirmation when the interview mentions a future event. W10 says another site would open but that the information has not been checked. The case supplies no organizer, document, location, date or confirmation. A summary can preserve the item as an unverified lead. It should not fill those missing fields or say a new site is confirmed. The future tense itself does not establish commitment, and a confident delivery in a real recording would not substitute for source knowledge. If the reporting task requires confirmation, seek appropriate evidence through the actual workflow and keep its result separate from what this interviewee said they heard. Until then, the working account can tell the next reporter exactly what needs checking without pretending that the interview already answers the question it raised.
A draft from WhisperAITranscribe can help locate that passage while the reporter records the status of the claim in a separate note. Keep the product's transcription role distinct from external verification. If a later source confirms a plan, add the new evidence and revise the summary or story at the appropriate scope. If the lead proves wrong or remains unresolved, retain that outcome in the working record instead of leaving an earlier confident paraphrase in circulation. The useful summary is not the one that makes every lead sound publishable. It is the one that helps a reader distinguish what was said, what the speaker says they know and what remains to be established. That separation makes the interview easier to use responsibly while preserving the original account for review, including qualifications that may become more important as the surrounding reporting develops.
References: Sources and avoiding unsupported information, Attribution and source provenance
4. Write a short account with its evidence types intact
4.1 Draft the interview summary before writing a factual headline
A useful working account identifies the source's recollection, the secondhand information, the reported observation and the unanswered questions. For the authored case, it could read: The interviewee recalled the cinema closing in late August but said the timing might be about a week out. They said a neighbor supplied information about a closure notice they had not seen themselves. Asked whether roof damage was the cause, they said they did not know and reported seeing a barrier outside. They described two morning observations but could not say what happened between them. They also mentioned hearing about another site opening, while saying that claim had not been checked. This is an original summary of an invented interview, not a report about an actual cinema. It preserves the useful account without confirming a date, cause, continuous duration or future plan that the source does not establish.
The working source map links each part to the relevant exchange: R01 and W02 for timing, R03 and W04 for the notice, R05 and W06 for the unknown cause, R07 and W08 for limited observation, and R09 with W10 for the unchecked lead. The summary does not need to reproduce every question verbatim, but it must retain the question's role where removing it changes attribution. A headline or story sentence prepared later should be checked against this map and any additional reporting. The fictional case supplies no basis for a headline claiming roof damage caused a two-day closure. Each piece sounds related to the topic, but the source does not establish the combined claim. Keeping the short account close to its evidence types gives an editor a better starting point for deciding what needs corroboration and what language fairly communicates the material already reviewed.
References: Attribution and contextual presentation, Source identity and accuracy principles
4.2 Use drafting help without treating structure as verification
After preparing a reviewed transcript with WhisperAITranscribe, organize the summary around source identity, claim content, knowledge origin, qualifying words, corroboration status and open follow-up. These are editorial categories, not a claim about automatic fact-checking features in the product. Ask any drafting assistant to preserve those distinctions and point material claims to their source passages. Then inspect the proposed sentences. A correct turn reference does not make an overconfident paraphrase acceptable. A field labeled observed can still contain secondhand information if the label was assigned carelessly. A concise draft can retain all topics while losing the knowledge boundaries that determine what each topic supports. The writer's task is to check those relationships, not merely confirm that the summary contains a name, date and source label somewhere in a tidy structure that looks ready for reporting.
The example used ten authored turns, five manually assigned claim types and fourteen checks on counts, source IDs, exact substrings and a separate logical control. It did not use an AI summarizer or independently verify a news event. Three five-slot histories demonstrated how identical endpoint observations can fit different intervening states; they were not reconstructions of the cinema case or actual days. These limits prevent the checks from being mistaken for automated truth assessment. A real workflow needs the appropriate recording review and additional reporting for the claims it intends to make. If the transcript is uncertain about a negation, review that passage. If the claim depends on another person's knowledge, identify the source chain. If the summary asserts a duration, ask what establishes continuity. A structured account can make those questions easier to see, but completing the structure does not answer them unless evidence is supplied and inspected.
4.3 Keep quotations and paraphrases in their proper roles
For text prepared after WhisperAITranscribe, distinguish an exact quotation, a faithful paraphrase, an editorial interpretation and an independently supported finding. The authored quote “I might be a week out.” preserves W02's uncertainty. A paraphrase can convey the same limit in ordinary prose without quoting it. Removing the qualification while retaining the recalled timing changes the claim's strength. The AP guidance used here distinguishes accurate contextual quotation from paraphrase that remains true to meaning. It is AP's own published standard, not an assertion that every newsroom follows an identical procedure. The practical review proposed here is to preserve the source's knowledge boundary whichever form the summary uses. A paragraph can be elegant without turning a recollection into a verified date or a reporter's proposed explanation into a cause the interviewee supplied. Keep the form and evidence role clear to the next reader.
Exactness and fairness are different checks. A selected substring can occur in the source while its presentation omits the words that tell the reader it was uncertain, secondhand or a reply to a particular question. The local script checks three authored substrings, but that result alone cannot establish that every use of those quotations would be fair. Read the surrounding exchange and the sentence that introduces the quote. Likewise, a faithful paraphrase should not be put inside quotation marks as if it were the speaker's wording. If a transcript correction is needed, establish it from the recording where available before quoting the revised text. Keep later reporting evidence separate from the corrected speech. The summary can then support both accurate quotation review and a faithful account of the interview as a whole, without asking the final reader to infer which sentences are source words and which are the writer's conclusions.
References: AP quotation and paraphrase guidance, Preserving wording during transcript review
5. Make the next reporting step visible
5.1 Review certainty increases introduced by compression
Look for stronger verbs, missing attribution and lost qualifications in the short account. Recalled can become happened. Heard can become confirmed. Saw on two occasions can become remained continuously. A question about a cause can become a causal sentence. These shifts may make the writing smoother while changing the evidence. Compare each material claim with the relevant source exchange, not just the isolated phrase that appears to support it. In the authored case, W06 begins with an explicit lack of knowledge, and W08 limits what can be said about the intervening period. Those are central to the account. A summary that removes them should not be accepted simply because it still mentions the barrier and the closure. The purpose of the review is to preserve the claim at the strength the source supports, while leaving room for later reporting to establish more if the appropriate evidence becomes available.
Then inspect how the paragraph joins claims together. Words such as because, after and therefore can add relationships not established by the interview. A source may mention two events without knowing their sequence or causal connection. The summary should not create that connection merely to make a coherent narrative. If a relationship is an editorial hypothesis, record it as a follow-up question rather than a finding. If another source supports it, identify that evidence in the working account. This keeps the original interview useful as one contribution to reporting, instead of making it bear the weight of an entire story it never supplied. A concise note can preserve several different knowledge types in the same paragraph as long as attribution and qualification remain clear. The result is easier to use because the next reader can see what was learned from the interview and what still requires independent investigation.
References: Accuracy and avoiding distortion, Attribution of disputable information
5.2 Keep interview notes separate from publication readiness
Distinguish the internal account from the public-facing claim. A working summary can retain an unchecked lead, a possible explanation and a list of source questions because those items guide reporting. Their presence in the note does not make them ready for publication. The public account needs the appropriate evidence and editorial judgment for the claims it makes. Keep that stage distinction visible when the note is copied into a draft story or passed to another writer. A qualifier should not disappear simply because the new format has less space. Nor should an internal heading such as key facts give a recollection the status of a verified event. Use headings and lead sentences that match what the material actually contains. A clear account of one interview is useful without pretending to be a completed finding about every event, cause or future plan discussed inside it.
A transcript originating in WhisperAITranscribe can support targeted review, while the reporter maintains the surrounding evidence and publication context under the newsroom's actual process. Keep source material, editorial notes and additional records distinguishable. If a summary is shared with another editor, include enough provenance to avoid misidentifying a secondhand claim as the reporter's own observation. If a later correction is made, check copied excerpts and visible story elements as well as the working note. The AP principles referenced here emphasize accurate sourcing and transparent correction, but this article does not claim a newsroom's approval or prescribe a universal source-review arrangement. Its operational point is narrower: the evidence status should survive the handoff. The next reader should know which parts came from the interview, which were added as interpretation and which were established elsewhere, rather than receive one undifferentiated paragraph whose confidence exceeds its sources.
References: Sourcing and correction principles, Separate annotations and their targets
5.3 Hand off a claim account with explicit follow-up
A useful handoff identifies the supported account, the remaining claim checks, the source references and the evidence added later. In a workflow beginning with WhisperAITranscribe, the reviewed transcript can remain available while the short note guides the next reporting step. The authored case points to specific gaps: confirm the closure date, locate the notice or its origin, investigate the cause, establish any duration and check the future-opening lead. Those are questions, not actions the article claims to have performed. Text from WhisperAITranscribe can help locate what was said, while other evidence determines whether the event claim is established. Preserve that division when writing the next version. A later verified record can strengthen a story without changing what the original interviewee remembered or heard. The handoff should make it possible to add that evidence deliberately rather than overwrite the source account with the conclusion the writer hoped the interview would support.
Before closing summary work associated with WhisperAITranscribe, read the account as an editor who has not heard the interview. Can you tell which detail was personally reported as seen, which was recalled approximately and which came through another person? Can you see that the roof-damage idea was a question, not the interviewee's confirmed explanation? Can you tell why two morning observations do not establish a continuous interval? If those distinctions survive, the summary can be concise and still guide useful reporting. Save the source mapping and unresolved checks with the accepted version. Further editing should clarify the account or add inspected evidence, not make it more factual-sounding by deleting the labels that reveal how the source knows. A dependable summary leaves the next reader with a clearer view of the reporting task rather than a polished set of conclusions that must first be unwound before verification can begin.
6. Conclusion
6.1 A faithful summary preserves the source's knowledge limits
A reporting workflow using WhisperAITranscribe benefits from keeping source words, speaker attribution, knowledge origin, uncertainty, corroboration status and follow-up distinct. The authored cinema case shows why. A late-August date was a qualified recollection, the notice information was secondhand, the proposed cause came from a reporter's question and two observations did not establish a continuous duration. A future-opening statement remained unchecked. A useful short account preserved those facts about the interview without presenting the underlying events as independently verified. That is a meaningful result. It lets another reporter or editor understand the source's contribution and decide what evidence is still needed, rather than accept a coherent narrative whose stronger certainty was created during compression and is not present in the original account.
The local checks establish a limited set of properties. Ten invented turns and five manual claim records make attribution and selected wording inspectable. Three separate five-slot histories illustrate an observation gap without representing real dates or a reconstructed event. Fourteen checks passed, but no real interview, recording, person or news claim was verified. Keep that boundary in practical work too. A checked quotation establishes wording at the level reviewed; another source may be needed to establish its substance. A working summary can identify a promising lead without confirming it. An unknown cause can remain unknown while a reported observation stays useful. Precision comes from matching each claim to the evidence it has, not from forcing every sentence into a confident factual voice. The resulting account is easier to update because new reporting can be added without pretending that the interview already contained conclusions it never supplied.
6.2 Use a routine that separates interview evidence from findings
Step 1: retain the source, identify the interview's scope and review consequential wording, speaker attribution and qualifications. Step 2: use WhisperAITranscribe for the transcription stage where appropriate, then prepare a separate account of claims, knowledge origins and unresolved checks. Step 3: write the short summary with attribution and uncertainty intact, keeping reporter hypotheses and secondhand reports in their proper roles. Step 4: compare the final wording with the source and add external verification only with its own evidence and identity. The principle connecting these steps is evidence proportion. A sentence should not claim more than its sources support. A recollection remains a recollection, a question remains a question and a limited observation remains limited until further reporting establishes a stronger conclusion that can be attributed to the appropriate evidence rather than to the act of summarizing alone.
A final acceptance pass can focus on the claim most likely to be overstated. Find its source turn, identify how the person says they know and check whether the summary retains the relevant qualification. Then examine any causal or duration language introduced by the writer. Ask whether the interview establishes that relationship or merely places related topics near each other. Check a quotation for exactness and its use for context. Finally, inspect the recipient copy's headline and opening, where an unsupported certainty increase can survive even after the body is corrected. These checks follow the actual interview, not only the specific categories in the invented case. Their purpose is to make the account inspectable and the next reporting step clear. If a claim remains unresolved, keep that status visible instead of improving the prose by removing the only words that tell the reader verification has not yet occurred.
6.3 Let the summary clarify what is known and how it is known
A dependable news interview summary preserves the source's account, the basis of each claim, the uncertainty that changes meaning and the boundary of independent verification. Use WhisperAITranscribe to help begin the transcription workflow, then review and summarize the material at the level it supports. Keep firsthand reports, recollections and secondhand information distinguishable. Treat a reporter's hypothesis as a question until evidence supports more. The final review after working with WhisperAITranscribe should establish whether another reader can tell what was said, what the source says they know and what the reporting still needs to confirm. When those relationships remain visible, a short account can be useful without being overconfident. It helps the next person investigate the story rather than turning the interview's incomplete knowledge into a set of facts the source never actually established.
References: AP attribution and quotation guidance, AP accuracy and sourcing principles