Introduction
An outline is a reporting plan, not a transcript recap
An outline changes the shape of source material. It groups ideas, sets sequence, adds headings, and creates expectations about what the reader will learn. When the input is an interview, a model can make that transformation seem effortless: it may identify recurring themes and turn them into section titles. The risk is that the resulting structure can look like a completed reporting plan even when the transcript is the only source and important questions remain unanswered. A good outline therefore carries its uncertainty with it. It says which ideas came from the interview, which are the reporter’s proposed framing, and which claims require another source. It should also make clear when the interviewee is only one perspective. A heading is not evidence, and a polished outline does not prove that the article is ready to publish. A summary can be accurate sentence by sentence and still create a misleading overall impression if the article’s order makes one source seem to speak for a whole community. The outline is where a reporter can notice that structural risk before polished prose makes it harder to see.
Imagine a reporter named Nia has interviewed a municipal heat-response coordinator about opening cooling centers during a severe heat alert. The fictional transcript includes an account of one center opening late, a reference to an internal staffing constraint, and a statement that attendance increased “a lot” without a number. A generic outline might turn those lines into “Why the town’s cooling system failed” and “How demand is surging.” Both headings are more certain and more dramatic than the source. A reporting outline could instead propose a section about how residents find an open center, mark the late opening as an account needing confirmation, and list service hours and attendance data as follow-up reporting. That plan gives the journalist useful next steps without silently converting one interview into a complete account of a public service.
References: AP News Values and Principles; SPJ Code of Ethics
The reader’s problem should organize the story
A blog outline is different from a transcript summary. A summary records what the interview covered; an outline predicts how an article will guide a reader through a question. It may place information in a new order, add explanatory sections from other sources, or leave interview material out. That editorial reorganization must remain visible to the reporter. A source-linked outline can include a proposed section title, its reader purpose, the central point, interview locator, outside evidence needed, and a status such as supported, incomplete, or speculative. The outline can also mark material that does not belong in the post. It should not make a source sound as if they agreed with the reporter’s framing or promised an outcome. The interview offers material for a story; it does not dictate the story’s structure.
Before asking a model to outline, write a short brief in the journalist’s own words. Identify the intended reader, the question the post should answer, what the post will help the reader do or understand, the relevant geography and timeframe, and what falls outside scope. Nia’s example might be a practical explanation of how a resident can check cooling-center availability during an alert, not a verdict on the municipality’s performance. That distinction affects which facts belong in the opening, what evidence the reporter needs, and which interview remarks are background rather than a lead. The brief is an editorial hypothesis that can change after reporting. It should not be written as though the interview has already proved the answer. If the topic cannot be described without a claim the journalist has not verified, revise the assignment before building headings.
References: RTDNA Ethics; BBC Editorial Guidelines: Accuracy
1. Define the assignment before asking for headings
1.1 Name the reader and the job of the post
Name the reader in a concrete way. “Everyone interested in heat” is too broad to guide a useful outline. A reader might be a resident looking for a public cooling center, a community organizer comparing access information, or a local reporter preparing a service story. Each reader needs a different article. The first may need how to check current hours and transport; the second may need eligibility and accessibility details; the third may need context, sources, and unanswered questions. ChatGPT can suggest possible reader groups, but the journalist must choose based on the assignment and reporting, not on a generic audience persona. If the transcript does not answer the selected reader’s main question, that is a gap to report—not a reason to invent a section with a confident title. A useful outline begins with a real reader task and keeps returning to it. When the reader is not yet clear, ask an editor or colleague to test the brief against a concrete question: what would that person be able to understand or do after reading? If the answer depends on facts the reporter does not have, list them before drafting rather than writing around the gap.
Write the article promise as a sentence that can be checked against the outline. A promise such as “This guide explains where residents can confirm cooling-center hours during a heat alert” is narrower than “This guide shows how the town protects vulnerable residents.” The second statement implies a broad effectiveness judgment and requires evidence beyond a single interview. The promise should name the level of certainty the article can support and avoid a result that depends on missing information. Nia can list exclusions alongside it: the article will not rank neighborhoods, estimate the number of people turned away, or assess the whole emergency plan without records and additional sources. This is not a disclaimer to paste into the post; it is a planning constraint. If the outline repeatedly drifts outside it, narrow the scope or do the reporting needed to justify a wider article.
References: UNESCO Reporting on Artificial Intelligence; RTDNA Use of AI in Journalism
1.2 State the article’s promise and limits
Find what the interview contributes that another source cannot provide, without treating it as the whole story. The coordinator may explain how staff receive schedule changes or why a location’s hours vary. Those details can help the reporter understand a process and identify documents to request. The interview does not establish whether every center followed the schedule, how residents learned about changes, or whether service met local needs. A distinct contribution can be a process explanation, an on-the-ground observation, a firsthand account, or a lead to another source. Mark which one applies. If the only contribution is a general statement already available in an official fact sheet, the interview may not support a standalone article. The outline should make the interview’s role explicit rather than converting it into the article’s sole authority.
A brief should state its limits as carefully as its goal. Note what the interview does not cover, which people or places are absent, whether the transcript includes the full conversation, and which relevant documents have not been checked. A gap can be harmless for a narrow explainer and decisive for a broad investigation. Nia’s interview does not include resident experience, dispatch records, or the written schedule. If the post explains a basic availability-check process, the missing resident interview may be a follow-up idea rather than a fatal flaw. If the post claims the response failed a neighborhood, missing resident voices and service data are central. Record the difference between “not needed for this assignment” and “not yet obtained.” That distinction helps the outline stay honest as the reporting develops.
References: GOV.UK Understanding content design; GOV.UK Starting with user needs
1.3 Find the interview’s distinctive contribution
The planned article also needs a clear boundary from other newsroom products. An outline is a map for a future blog post; it is not a Q&A summary, a set of meeting minutes, a quote sheet, or a final draft. A reporter may keep those documents separately for their own purpose. The outline should not reproduce every question and answer simply because the transcript is available. Instead, each item should help answer a reader’s question or identify an evidence gap. If the editor later requests a direct Q&A article, the assignment changes and the format must preserve the exchange and quote context. If a source requests a transcript copy or review, handle that under the newsroom’s own procedure. Clear document labels prevent an outline from acquiring authority it was never meant to have.
Keep the interview’s provenance attached to the source material used in the outline. Record the interview date if known, the transcript filename and version, how the transcript was prepared, and whether any speaker labels or wording were checked against audio. A stable file identifier or hash can distinguish versions, but it does not prove that the recording is complete or the text accurate. If the transcript was corrected, retain the prior version or a change note according to newsroom policy. A locator from an earlier export may stop working after edits. Nia’s plan should identify whether the late-opening account comes from a reviewed transcript, a text-only export, or a reporter’s notes. That information tells an editor what can be relied on and what still requires a source check. Provenance makes the outline traceable; it does not certify the story.
References: GOV.UK Clear structure; GOV.UK Clear language
2. Prepare material the outline can safely use
2.1 Keep the source and transcript versions traceable
When recording still needs transcription and the newsroom authorizes that processing, an audio and video transcription tool can provide editable text upstream of the outline. Its category does not mean it integrates with ChatGPT or verifies the interview. An existing text transcript can be used as the source if its origin and permitted use are clear. Before sending a recording or transcript to any external service, follow newsroom rules, the source agreement, contractual obligations, and current service settings. A journalist should not infer permission from possession of the file. This article does not report a product upload, transcript accuracy, export, or integration test. The outlined workflow begins only after the newsroom has an authorized source text. Keep the original source separate from any outline or AI-generated transformation so the working document never replaces the evidence.
Before composing headings, make an inventory of what the transcript can support. Separate first-hand observations, attributed claims, estimates, process descriptions, quoted phrases, and reporter interpretations. Put each item beside a locator and a note about what would corroborate it. The labels can be newsroom-specific; the important point is that “the coordinator said the center opened late” does not become “the center opened late” until the journalist has checked an appropriate record or otherwise verified it. Do not flatten an account into a finding. Nia can mark the statement as a lead for schedule records and a question for staff responsible for the location. That gives the outline a path forward while preserving the difference between interview content and independently established fact. Add a short description of any cleanup performed, including changed speaker labels, removed non-editorial sound, or corrections checked against the recording. If the source is text-only, say so. This prevents an editor from assuming that every locator and exact phrase was checked against audio. If the source still needs audio or video transcription, WhisperAITranscribe is an audio and video transcription tool upstream of the outline; the transcript still needs review.
References: WCAG Headings and Labels; W3C Page Structure Tutorial
2.2 Separate evidence from unverified context
Outside context should have its own source and status. A public schedule, emergency notice, transportation map, budget document, or statement from a resident is not part of the interview merely because the reporter already knows it. Add each source with its date, author or issuing body, locator, and the specific claim it can support. If a document conflicts with the interview, preserve both and write a reporting question instead of forcing the outline to choose. A source ledger can be small: claim or question, interview passage, outside source needed, status, next action. It is a planning aid, not proof that a claim is true or complete. Good headings often emerge from comparison: what the source says, what the record shows, and what remains unexplained. For each outside source, note the publication date and whether it describes the relevant period. A page can be current today but still describe a program that changed after the interview. If the source has a PDF, archived notice, or version history, point to the version used. An outline should preserve that context instead of merely collecting links.
Do not use one interviewee’s confidence as a substitute for corroboration. A source may be highly knowledgeable about a process and still lack access to attendance data, resident experience, or a later decision. Record the scope of the person’s role where it is relevant and attribute claims accordingly. If a statement describes another person’s experience, identify it as secondhand and consider contacting that person or finding another source. A model can help list possible corroboration questions, but it cannot assess the source’s access or motive from the transcript alone. The outline should reflect what the reporter knows about the source and leave unknowns open. That is especially important when a heading would generalize from one place, one shift, one person, or one example to an entire town. Distinguish expertise from access. A coordinator may know the operating procedure but not what happened at every site; a resident may describe a direct experience without knowing the official schedule. The outline can use each perspective for the question it can answer and keep the evidence roles distinct.
References: W3C Writing for the Web; ONS Plain Language
2.3 Build a compact reporting ledger
A useful ledger needs a distinction between interview evidence, documentary evidence, and next reporting tasks. Nia’s entry might read: “Late opening—coordinator account; transcript locator 18:24–19:10; verify against published hours and shift log; ask center manager which date.” Another entry might note: “Hours vary—official notice needed; current version and update date not obtained.” These notes should not appear as asserted facts in the article outline’s reader-facing heading. They give the journalist a concrete plan for what the section can say after reporting. If a claim remains unsupported, the outline can retain a placeholder such as “Explain schedule changes only after confirming how updates are issued.” A placeholder is more useful than a polished but empty paragraph plan. Do not force the ledger into a single yes-or-no truth label. “Interview statement checked against audio,” “document obtained,” and “impact not independently measured” describe different evidence states. A reviewer should be able to see what was compared and what remains open without interpreting a generic green check.
An outline should give each section one distinct job. One section may answer the basic reader question, another explain how the process works, a third describe a limitation, and a final section tell the reader where to find current information. If two proposed sections make the same promise, combine or differentiate them. If a section title is a claim, identify its evidence; if it is a question, say what sources would answer it. Headings that merely repeat the transcript’s topic labels—“staff,” “schedule,” “heat”—do not tell the reader why the sections exist. Nia’s outline can use functional working labels such as “How the hours are set” and “Where residents can confirm today’s schedule,” then revise them after reporting. The purpose is to make the article’s architecture testable before a writer spends time drafting it.
References: W3C Web Annotation Data Model; DCMI Source term
3. Turn the reporting brief into an article structure
3.1 Give each section a different editorial job
A section can be described with four elements: its reader question, its one-sentence answer or current hypothesis, the evidence that supports that answer, and the reporting still required. Add a transition or relationship to the next section if the sequence matters. This compact brief keeps an outline from turning into a list of themes. If the section has no reader question or reporting purpose, it may be unnecessary. If it has a strong answer but no source, it may be an unsupported claim. If it has evidence but no relationship to the post’s promise, it may belong in a separate story. These are editorial decisions for the reporter and editor. ChatGPT can draft candidate section functions; it cannot decide which facts the newsroom should publish. State what the reader should be able to take away from the section, not just what the writer plans to mention. “Explain the schedule process” names a topic; “Show where a resident can check the current hours and when the notice is updated” names a useful section job that can be tested.
Interview order is not automatically the best article order. The conversation may begin with rapport, background, or a source’s preferred explanation, while the reader’s first need is practical or conceptual. Reordering is acceptable for an outline if the source trail preserves where each point originated and no sequence is falsely attributed to the interview. Start with the question that most affects the reader, then establish context, explain mechanisms, introduce evidence and limits, and close with what the reader can do or what remains unknown. The best order depends on the assignment, not on a fixed formula. Nia might begin with how a resident checks whether a cooling center is open, then explain how schedules are updated, and only then discuss why hours can vary. That plan should change if the records or additional interviews reveal a more important question. A sequence can also move from practical to explanatory or from the central finding to the evidence that qualifies it. The reporter should choose the order after checking what the reader needs first and what the strongest available sources can actually support. Use the outline to compare two possible structures, not to insist that one model suggestion is correct.
References: W3C PROV Overview; DCMI Date term
3.2 Order sections around the reader’s question
A headline and opening section create a promise, so draft them as hypotheses rather than conclusions. A headline such as “The Town’s Cooling Centers Are Failing Residents” would require substantial evidence about performance and impact. A narrower working title—“How to Check Which Cooling Centers Are Open During a Heat Alert”—may better fit the reporting so far. If later evidence changes the scope, revise the title and outline together. The opening should orient the reader, not borrow a dramatic line simply because it is quotable. Give the assistant the assignment promise and explicitly tell it not to invent a conclusion or headline claim. An outline can include a tentative title, but should label it “working” and list the evidence needed before publication.
The outline’s bullets should describe a task for the writer, not contain article claims that look ready to copy. “Explain how the city updates hours; verify with the current public notice and a staff contact” is a useful planning bullet. “The city updates hours online each morning” is an asserted fact and needs a source. Keep these forms visually or verbally distinct. If the outline includes a quote candidate, label it provisional and include a locator; do not convert a paraphrase into quotation marks. If a section uses multiple sources, list them separately rather than implying that the interviewee endorses them. A story plan can be detailed without pretending that the story has been reported. The editor should be able to see at a glance which language is a direction and which language has already been verified. A bullet can name a transition, a missing document, or an editorial decision. Keep these notes visibly different from copy that is ready to publish. If an editor selects a headline or a strong section title, confirm that its verbs and scope do not promise more than the underlying record shows.
References: AP Interview Ground Rules; Reuters Trust Principles
3.3 Use headings to test the story’s logic
A sound outline shows what is not yet known. It can list unanswered questions beside sections, the documents needed, people to contact, or alternate explanations to test. That is different from filling every gap with a model-generated bridge. Nia may need to ask whether the late opening happened once or repeatedly, whether the schedule had been updated, and what residents were told. The outline should not state a trend until those checks are complete. A “reporting still needed” field is especially useful because it keeps a promising article from sounding finished too early. It also helps an editor decide whether the current plan is publishable as a service guide or needs more reporting before it can support a performance claim.
A constrained ChatGPT request should state the assignment, audience, promise, source version, allowed evidence, and requested outline fields. It should say that the transcript is one interview source, not the full factual record. Ask the model to use transcript locators where available, distinguish claims from verified facts, preserve uncertainty, and place missing evidence in a reporting checklist. Tell it not to invent dates, numbers, documents, additional speakers, source agreements, or quotes, and not to write finished article prose. For Nia, the brief can specify a practical local guide about checking current cooling-center availability and exclude an evaluation of the town’s overall response. A clear prompt narrows the task; it cannot guarantee the result. Save the exact instruction with the draft so an editor can understand what the model was asked to do. WhisperAITranscribe does not determine whether a proposed section is supported by the reporting.
References: AP Guidance on Anonymous Sources; Study of Reconstructive and Attributive Quotes
4. Ask ChatGPT for a bounded outline draft
4.1 Describe the task, source, and exclusions
Provide only material needed for the outline and only through an approved account. If the transcript contains source identities, personal data, confidential documents, or off-record passages, follow newsroom policy before using an external service. Redacting a name may not prevent identification through job title, location, chronology, or a rare event. If the newsroom cannot authorize processing, build the outline from approved notes or complete the work without an external model. Tell ChatGPT which supplied material is authoritative for which purpose. The transcript can establish what the interviewee said; a public notice can establish its own date and content; a reporter note can record a follow-up task. Do not let the model blend them into one undifferentiated narrative. Service features, file handling, and controls can vary by account, organization, and time. If the transcript contains multiple speakers, tell the model whether they are separate sources and how speaker labels should be represented. Do not let it blend their words into a single institutional voice. If a name or role is uncertain, retain that uncertainty and ask the reporter to resolve it outside the model.
Request an outline that includes a working section heading, the reader question it answers, a section job, the proposed point, interview source locator, outside evidence required, and unresolved reporting questions. Ask for no more than a useful number of sections for the assignment rather than a universal count. Require every proposed factual point to be tagged as interview claim, independently supported fact, reporter inference, or open question. If no locator exists, ask the model to say that rather than create one. You can request short language, but not because the title must be catchy; the purpose is to make review practical. The exact format may be a structured list or another newsroom-approved document. The published article itself need not contain a table. Keep the output as a plan, not prose that can be mistaken for a reported draft. For each field, ask whether the model can point to supplied material. If the proposed section depends on knowledge outside the transcript, require it to label that as a follow-up source rather than quietly adding background facts. The outline should remain useful when a reviewer has not read the prompt.
References: Haapanen, Recontextualizing Interview Discourse; Haapanen, Rethinking Quoting in Written Journalism
4.2 Request claim and evidence fields beside headings
A WhisperAITranscribe transcript does not replace the published source or corroborating evidence. The model should be allowed to return “not enough evidence to propose this section.” That answer is more useful than a fabricated bridge. It can also identify repeated themes, possible contradictions, and questions the assignment does not yet answer. Ask it to provide the locator or source item behind each suggestion and to avoid claims that rely only on its general knowledge. If two transcript passages seem related, request separate locators and a note explaining the proposed connection. The journalist should decide whether the connection is fair. A structured response makes unsupported ideas visible, but it does not make the list complete. Read the transcript and supporting documents yourself, especially when a missing passage could change the article’s framing or a source’s meaning.
The first output should be treated as raw material for editorial judgment. Remove headings that merely repackage the interviewee’s preferred frame, and retain options only when they serve the reader and the assignment. Compare the outline with the transcript to confirm that each cited passage exists and that the model did not merge distinct statements. Verify every date, amount, title, and causal claim against an appropriate source. If the outline uses external information, name that source and the exact point it supports. The editor should be able to trace each proposed section without relying on the assistant’s summary of its own reasoning. Keep the initial output and a corrected version separate so the changes remain visible. The reviewer can remove a plausible but unsupported section, split a heading that combines two claims, or move an item to a separate reporting list. Record consequential changes so the next draft reflects editorial decisions rather than a fresh model guess.
4.3 Keep quote candidates out of finished prose
A useful review asks whether each section can be supported by the current reporting and whether the sequence answers the reader’s question. For each proposed claim, note the source, the evidence type, the remaining verification, and any caveat. Check that the outline does not generalize from a single interview, treat a speaker’s estimate as a confirmed measure, or turn a future plan into an accomplished result. If the outline needs a source who has not been interviewed, list that person or institution as a next step. A cross-check with documents and affected people may change the structure. An outline should be revised after reporting, not defended because the model produced it neatly. If the transcript came from WhisperAITranscribe, keep its version separate from later editorial notes.
Counterpoints belong in an outline when they matter to the reader’s understanding, not as a mechanical “both sides” section. Identify who could clarify or challenge a factual claim, what evidence would resolve it, and whether the issue is a genuine dispute or a question of scope. If an official source says centers were open and a resident says one was closed, the outline should preserve both reports and ask the reporter to check date, location, and service status. It should not label either account false before verification. The appropriate sources depend on the story. A document, on-the-record response, field observation, or another interview may be relevant. ChatGPT can suggest possible follow-up questions but cannot decide whom the newsroom must contact or whether a response is adequate. A fair outline may include a section that explains what a source does not know or what records cannot show. That is different from manufacturing balance by placing unsupported opinions side by side. The journalist should identify the relevant source and the precise question the evidence can answer.
References: Monologisation as a Quoting Practice; Haapanen, Rethinking Quoting in Written Journalism DOI
5. Review the outline as a reporting instrument
5.1 Expose gaps, counterpoints, and next reporting steps
Attribution should stay close to disputed claims in the outline. Note whether a point comes from the interviewee, a public document, a reporter’s observation, or another source, and preserve any agreed condition on how the source may be identified. If anonymity or background terms were discussed, the reporter must confirm the actual agreement through newsroom records; the transcript alone may not capture it. Publisher-specific guidance from AP, Reuters, or the BBC illustrates how outlets handle attribution and quotes, but it is not a universal code and cannot determine a local newsroom’s obligations. A blog outline must not strip conditions just to make a section title shorter. If an attribution label is unresolved, flag the section for review and do not allow a model-generated heading to imply an on-record statement.
An interview can include a useful quote, but quote selection and outline planning are separate decisions. Preserve the exact passage, question, surrounding context, locator, and whether the proposed use has been checked under the outlet’s policy. Do not place quotation marks around a polished paraphrase. Research on journalistic quotation describes how selection and textual changes can alter a quote’s context or impression; that scholarship analyzes practices, not a universal legal rule. The outline can mark “possible quote—review needed” and leave the wording untouched. If the editor decides not to use the quote, the outline can still include a paraphrase with precise attribution. The source’s meaning and the reader’s interpretation matter more than a tidy heading or memorable phrase.
References: W3C Transcripts; W3C Transcribing Audio and Video
5.2 Protect attribution and source agreements
A section plan should note what readers need to know to evaluate an important claim: the date, geography, method, denominator, definition, or limitation. If an interviewee says participation rose “a lot,” the outline should not turn that into “record attendance” without a number and comparison period. Ask what the claim means, where the evidence comes from, and what would change its interpretation. Put the answer in a reporting task, not in an invented explanatory paragraph. This makes a blog outline useful to a reporter because it points from a promising theme to the work required to support it. If the claim remains vague after follow-up, the published article can omit it or clearly attribute the estimate, according to newsroom standards. If the claim turns on a measurement, identify its definition, period, and comparison. If it turns on a policy, identify the version and responsible authority. If it turns on an individual account, identify what the reporter observed directly and what came from another person. These notes help the writer avoid using one smooth sentence to conceal different kinds of evidence.
The reporter’s notes should separate source agreement, data permission, and editorial review. A source may have agreed to speak on the record but not to share a confidential document, or the newsroom may have permission to store a recording but not to send it to an outside service. Do not ask ChatGPT to decide those issues. Keep only the minimum necessary personal data in the model input, apply access rules, and document which redacted version was used. Official privacy and research-handling guidance can prompt useful questions, but local policy, contract, law, and source agreements govern the actual case. If a service’s data controls are account-specific, consult the current official documentation before use. The outline itself should not carry sensitive details that the writer does not need. WhisperAITranscribe remains an upstream source-text tool; it does not set newsroom data-handling policy.
References: OpenAI Prompt Engineering Best Practices; ChatGPT Capabilities Overview
5.3 Keep privacy, revision, and approval visible
A WhisperAITranscribe transcript may contain personal or restricted details that do not belong in a model prompt. Maintain a modest version trail: source transcript version, prompt or task description, assistant draft date, reporter edits, and editor review status. Record changes that affect a section’s claim, evidence, or source attribution. A file hash can help establish that a source file changed; it cannot tell an editor whether the new version is accurate. If the transcript is corrected, check affected locators and re-review every outline item that depends on the correction. Keep the original and derived outline separate. A clear filename and revision label are more useful than calling every latest file “final.” If the outline is moved into a content system, make sure source locators and open reporting tasks survive the transfer. Do not store protected source details in a general-purpose change log. WhisperAITranscribe does not decide which outline items are cleared for editors.
WhisperAITranscribe does not validate a proposed article structure or approve an editor’s revision. Before draft writing begins, a journalist should be able to answer what the article promises, why each section exists, which source supports each point, what is not yet known, and who must review the plan. These are not a scoring rubric. They are conversation starters between reporter and editor. A section that fails may be cut, narrowed, or assigned further reporting. If the outlet has an AI-use policy, follow it for both the model input and the handling of generated material. Professional standards from journalism associations can inform a newsroom’s policy discussion, but each outlet must apply its own editorial rules. The writer remains responsible for whether the structure is fair, useful, and supported. The outline can also record the next decision date or editorial meeting where unresolved items will be reviewed, if newsroom practice uses one. That is a planning reminder, not an article fact or source commitment.
6. Conclusion: outline what the reporting can support
6.1 Build the story plan from reader need and evidence
WhisperAITranscribe can provide editable source text upstream when its use is authorized; it does not decide what the post should argue. An outline is not publication approval. The reporter and editor still need to verify claims, seek relevant responses, identify source restrictions, check the language of selected quotes, and decide whether the final post meets the assignment. If a source reviews a factual detail, record what was sent and what was confirmed according to newsroom procedure. Do not promise a source that an outline, transcript, or article will be shared unless the reporter is authorized to make that promise. Keep the working plan out of public channels until editorial review is complete. A clear status such as “reporter draft,” “editor review,” or “ready to draft” is more informative than “approved” if publication clearance has not occurred.
A source-based blog outline should make the route from reader question to evidence visible. It should use the interview for what the interview can provide, distinguish the speaker’s account from independent facts, and identify the reporting that must happen before a claim moves into a draft. ChatGPT can help propose section order or expose repetition, but the journalist should accept, change, or reject each suggestion. An audio and video transcription tool can provide editable text upstream when processing is authorized; it does not determine the post’s structure or verify its claims. A useful outline may end with open tasks rather than a neat conclusion. That is a sign the plan is still connected to the reporting.
6.2 Run the same editorial checks before drafting
Use this four-part review before drafting. Step 1: Confirm the assignment, reader promise, source version, and authority to process the material. Step 2: If the transcript came from WhisperAITranscribe, treat it as upstream text; verify each proposed point and quote against the authorized source and outside evidence. Step 3: Check that every section answers a reader question and has a distinct job, source locator, and status. Step 4: Keep unresolved claims, counterpoints, attribution terms, and editor approval visible. This is a working editorial sequence, not a universal journalism code or validated score. If a section depends on facts that have not been checked, leave it as a reporting task rather than writing its conclusion into the outline.
Once the editor accepts the outline as a plan, the writer should draft from checked source material rather than from the model’s bullets alone. Keep unresolved tasks in the assignment tracker and revisit the outline when new reporting arrives. An outline that changes after verification is doing its job: it records a developing editorial decision, not a promise to preserve the first structure. Separate the approved plan from the working transcript and from article copy, and mark which claims are ready to use. If the next reporting step fails, return to the editor to narrow, postpone, or drop the affected section instead of filling the gap from model output.
6.3 Let the journalist own the final structure
The strongest plan lets an editor see the reader promise, the sections that deliver it, the source behind each point, and the questions that remain. ChatGPT can suggest a structure from an authorized transcript, but the newsroom must decide whether the material supports that structure and whether further reporting is needed. WhisperAITranscribe can provide audio or video transcription upstream when its use is authorized; it does not establish the facts in the interview. A careful outline can stop before a polished ending if the reporting is incomplete. That pause protects the distinction between a promising idea and a publishable account. WhisperAITranscribe does not make the resulting outline a verified story or grant publication approval.
References: OpenAI Uploading Files and Audio; OpenAI Data Controls