Draft a Story Outline From Verified Reporting With AI at the Cursor

You can draft a story outline from verified reporting by speaking a short set of structured inputs at the cursor: each verified fact, its source, the evidence still missing, and the next question. An AI request then arranges them in a compatible Windows text field. The instruction that matters most is the constraint: use only these inputs, flag gaps, and invent nothing. The result is a working structure with the holes marked. It is not a story draft. You review every line against your own notes before the outline guides any writing, editing or conversation with an editor.

This article covers how to set that up. It includes an illustrative worked example, a reusable spoken template, a review checklist, a reference table and a method for measuring whether the workflow helps you.

A story outline does several jobs at once. It shows what you can already say, what you cannot yet support, and how the material might be sequenced. Much of this often lives in notebooks, message threads and half-finished documents. Turning it into a structure takes effort, and gaps can quietly disappear along the way. A fact that was “probably confirmed” late yesterday can show up in the outline as settled.

The aim is an outline that keeps gaps visible. Every entry carries its source beside it, and every weak spot becomes a question with a person attached. That makes the outline a reporting tool as well as a writing tool, because it tells you what to chase next.

The review boundary is firm. The AI arranges what you give it. It does not report, verify or supply quotations, and nothing it produces should be treated as something a person said. This workflow is not about recording or transcribing interviews. It covers your own spoken inputs and instructions. Any quotation in your story must come from your verified notes or recordings, handled however your newsroom requires.

The quality of the outline follows the quality of the inputs, so settle these four fields first, in your head or on paper.

1. Verified fact. A claim you have confirmed, stated plainly and narrowly. “The council voted 5–2 on Tuesday” is a verified fact. “The council seems to support the plan” is an interpretation and belongs elsewhere.
2. Source. What or who confirmed it: a named document, a named person on the record, a dataset, a filing. Include the type of access, such as on the record, background or document only, so the outline preserves it.
3. Missing evidence. What would strengthen or complicate the fact: a second source, a document you have not seen, a figure you have not checked, a response you are waiting for.
4. Next question. The specific question you will ask, and of whom, to close the gap.

Keep each entry short. An instruction with six or seven well-formed entries is easier to review than one with twenty. For a large story, work in batches by theme, such as background, decision, impact and response, and generate one section at a time.

Syping is speaking plus typing: you speak, and text appears where you would otherwise type. In SpeakLingo this covers online transcription of your speech in 99+ languages, and spoken AI requests that produce text at a compatible focused Windows text field. AI contextual formatting helps shape that output into a usable structure, such as headings, bullet points and labeled gaps. The features page gives the wider product overview, and the use cases page shows how other kinds of work map onto it.

For outlining, the practical appeal is staying in the document. You place the cursor in your outline file or notes field, speak the instruction and the four-field inputs, and the generated text lands there. Where the supported cursor workflow fits your setup, you may avoid opening a separate AI window, pasting material in, copying the result and pasting it back. You still do the structuring judgment yourself. The tool reduces manual arranging.

Some limits shape how you use it:

• Compatibility varies. SpeakLingo is not universally compatible with every Windows application or field. Test your usual outlining tool well before a deadline.
• It does not act on its own. It is not an autonomous agent that operates your computer, files stories or publishes.
• AI requests have requirements. They need an internet connection, an enabled service, your consent and the relevant access.
• Data may leave your device. Audio and text may be sent to remote services to process a request, and optional screen context may be sent if you opt in. Read the consent prompts and decide whether that suits the material.
• AI output needs review. The outline is a draft by definition.

Do this once, with throwaway content, before you rely on it for real reporting.

One boundary matters before you start. Dictating alone does not turn on AI. Use the supported AI Task workflow, and put the instruction and every fact it needs (each verified fact, its source, the missing evidence and the next question) in that same spoken request. Your earlier dictations, text already in the outline field and whatever is on your screen are not automatically available to it. Screen context is a separate opt-in, and it may be sent to a remote service. Instruct the AI to limit its output to the facts in your request, but do not assume it will comply perfectly. It may still add details you did not supply, so review every line and treat anything beyond your inputs as unverified until you confirm it against your notes.

1. Choose a field and test it. Pick the document or notes field where your outlines live. Speak a harmless test sentence and confirm text lands where the cursor is. If the field does not respond properly, try another one.
2. Confirm the AI requirements. Check that you have an internet connection, that the service is enabled for you and that you have given consent. Decide in advance whether the material in front of you is suitable to send to a remote service, and whether you want to opt in to screen context at all.
3. Turn off automatic Enter or send behavior. SpeakLingo has an optional setting that can press Enter after text is delivered, and in some apps that can submit or send what you spoke. For this workflow, switch it off, especially if your outline field sits inside a chat, email or publishing tool. Nothing should be submitted before you have read it.
4. Speak the instruction, then the inputs. Use the template below so the constraint comes first and the gaps are requested explicitly.
5. Review before using. Read the generated text against your notes while it is still a draft in your own document. Only then should it guide reporting, go to an editor or move into another tool.
6. Keep the open items. Leave the gap lines in place until each one is closed. They are the most useful part of the output.

This example is fictional and uses non-sensitive details. The town, street, people and figures are invented to show the method. The “after” text illustrates the kind of structure an AI request might produce. It is illustrative, not observed output from any product test.

The story: a fictional local reporter is covering a proposed protected bike lane on Quarry Road in the fictional town of Harbor Mill.

Before: the scattered state

The reporter’s notes sit in three places. One says the council voted on the lane. Another says a “business owner was upset.” A third holds a half-remembered budget figure. An outline typed in a hurry reads:

Council approved bike lane — big win for cyclists. Budget about $2M? Businesses angry. Traffic study says it's fine. Need quotes.

The problems are easy to list. “Big win” is a judgment. The budget figure has a question mark and no source. “Businesses angry” generalizes from one conversation. “Traffic study says it’s fine” names no study and does not say what “fine” means. The outline looks complete while hiding every weakness.

The spoken instruction

At the cursor in the outline document, the reporter speaks something like this:

Using only these verified inputs, draft a story outline from verified reporting. Use a clear professional structure. Flag missing facts; do not invent them. Do not write any quotations.

Input one. Verified fact: the Harbor Mill council voted 5 to 2 to approve a protected bike lane on Quarry Road. Source: the posted meeting minutes. Missing evidence: the minutes do not state the cost. Next question: ask the public works office for the approved project cost and funding source.

Input two. Verified fact: one owner of a shop on Quarry Road told me she is concerned about loss of curbside parking. Source: my phone interview with her, on the record. Missing evidence: no other business owners contacted yet, so I cannot say what businesses in general think. Next question: ask the local merchants' association whether it has taken a position, and contact at least two more shop owners.

Input three. Verified fact: the city commissioned a traffic study on the proposal. Source: reference in the agenda packet. Missing evidence: I have not read the study itself. Next question: request the full study and ask the author what it measured.

After: the outline for review

The generated text at the cursor might look like this before the reporter edits it:

Working headline direction: Harbor Mill council approves protected bike lane on Quarry Road

1. The decision - Council voted 5–2 to approve a protected bike lane on Quarry Road. Source: posted meeting minutes. - Gap: Project cost and funding source not stated in the minutes. Next: Ask public works for the approved cost and funding source.

2. Local reaction - One Quarry Road shop owner is concerned about loss of curbside parking. Source: on-the-record phone interview. - Gap: Only one business owner contacted; no basis yet to describe broader business opinion. Next: Contact the merchants' association and at least two more shop owners.

3. The evidence behind the plan - A traffic study was commissioned. Source: agenda packet reference. - Gap: Study not yet read; findings unknown. Next: Request the full study and ask its author what it measured.

Open items before drafting: cost and funding, broader business response, study findings. No quotations included.

The reporter then reviews it line by line against the notes. In this illustration, the headline direction stays provisional until the cost question is answered. The “Local reaction” heading is renamed to reflect that only one voice is in hand. The reporter also adds a line about who has not yet been asked, such as cycling advocates and residents. The outline now shows what the story can support today and what it cannot.

Adapt this to your own voice and your newsroom’s habits, and speak it at the cursor in your outline document.

Using only these verified inputs, draft a story outline from verified reporting. Use a clear professional structure with a section for each theme. Under each point, show the source in italics. Mark every missing piece of evidence as a gap and pair it with my next question. Do not invent facts, names, figures or quotations. Do not write any quotation. List open items at the end.

Story topic: [one sentence].

Input [number]. Verified fact: [claim, stated narrowly]. Source: [document, person, access type]. Missing evidence: [what is not yet confirmed]. Next question: [question and who to ask].

[Repeat for each input.]

Anything I have not listed is unverified and should not appear as fact.

The last line gives the AI an explicit rule for everything outside your inputs. It will not prevent every embellishment, so the review step still matters, but it sets the right expectation.

This checklist is an illustrative writing practice, not a newsroom standard or professional advice. Adapt it to your own newsroom's rules, and run it on every AI-assisted outline before it guides your reporting or goes to anyone else.

• [ ] Every fact traces to my notes. I can point to the document, interview or dataset behind each line.
• [ ] No invented material. No names, numbers, dates, titles or places that I did not supply.
• [ ] No fabricated quotations. Nothing appears in quotation marks unless I placed it there from verified material.
• [ ] Attribution preserved. On-the-record, background and document-only distinctions match what I agreed.
• [ ] Interpretation labeled. Phrases such as “big win” or “widely opposed” are removed or flagged as unsupported.
• [ ] Gaps still visible. Each missing item is present, accurate and paired with a next question.
• [ ] Scope matches evidence. One source is not described as “residents” or “businesses.”
• [ ] Balance noted. I have listed who has not yet been asked, including people affected or criticized.
• [ ] Names and numbers checked. Spellings, figures and titles match the originals, including anything speech transcription could have misheard.
• [ ] Sensitive details decided. I have chosen whether any source-identifying detail belongs in this document at all.
• [ ] Nothing sent early. Automatic Enter or send behavior was off, and I read the text before it went anywhere.
• [ ] Editor-ready. If I share this, the gaps and sourcing are clear enough that a colleague would not mistake it for a finished draft.

Names and figures deserve extra attention, since a misheard word is easy to miss and costly to carry into copy.

This table maps the parts of an outline to what you supply, what to check and how to mark uncertainty.

Text appearing quickly at the cursor tells you little. The useful measure is time to reviewed output: the time from starting the outline to holding a version you have checked and would stand behind.

1. Pick comparable tasks. Choose several stories of similar complexity, such as routine meeting or court stories, rather than a one-off investigation.
2. Start the clock at the first input. Stop it when you have finished the checklist above and the outline is ready for use.
3. Record the steps. Note the time spent speaking inputs, waiting, correcting and checking, including any time fixing invented or misplaced material.
4. Compare with your usual method. Time a few outlines built manually, with the same start and stop points.
5. Note quality alongside minutes. Did the outline keep gaps visible? Did you catch more or fewer problems later in drafting?
6. Repeat across several stories. A single attempt tells you little, since a tricky topic or a poor connection can skew it.

If review and correction take as long as building the outline by hand, the workflow is not helping that kind of story, which is a useful finding. Your own log is the relevant evidence.

Journalism carries obligations that most office work does not. Audio and text may be sent to remote services when you use online transcription or AI requests, and optional screen context may be sent if you opt in, so think carefully about what you speak. For confidential sources, unpublished documents or anything where exposure could cause harm, decide whether to use the tool at all. That decision is yours, alongside any newsroom policy you work under.

One illustrative writing practice is to describe sensitive inputs generically when speaking, such as “a source with direct knowledge,” and keep identifying details in your own secure notes. Whether that suffices depends on your situation, your organization’s rules and your judgment. This article is not legal or professional advice, and it should not be read as a guarantee about confidentiality or data handling.

If your internet connection is interrupted, offline English transcription is only a fallback, and only if you have already configured it, for getting words down. Transcription has its own mode, setup and access requirements, so AI being unavailable does not show that transcription is available. AI requests do not work offline, so the outline would have to be built by hand or after you reconnect.

Online transcription supports 99+ languages, so you can speak your inputs in a supported language that suits you. That is transcription, not translation, and it does not guarantee AI output in every language. If you report in one language and file in another, handle translation yourself and review carefully, especially names, titles and anything attributed to a source speaking a different language.