What a Local News Website actually does
Summarizes meeting transcripts, drafts newsletter sections, generates SEO headlines, and writes public-records request templates so a 1–6 person news operation can cover more beats without adding staff.
A hyperlocal news operation's constraint is reporter time. A single beat reporter covering city hall can spend 2 hours transcribing and summarizing a 90-minute city council meeting before writing a single sentence of journalism. Otter.ai at $16.99/mo handles the transcription; Claude Sonnet 4.6 converts the transcript into a structured meeting summary in 15 minutes instead of 2 hours. That 4–6 hours recovered per week is the difference between covering one beat and two.
In 2026, local news is experiencing a subscription revenue inflection: the Local Media Association reports that hyperlocal newsletter publishers with 1,000–10,000 subscribers are achieving 8–15% paid conversion rates — higher than national outlets. Retention, not acquisition, is the primary metric. AI helps on the operational side of retention: newsletter formatting, subscriber win-back sequences, SEO headline variants, and public-records request templates that increase the volume of documents a reporter can request without increasing research time. The critical line that cannot be crossed: AI must never write the actual reporting, generate quotes attributed to sources, or publish content without human verification. One fabrication ends a local news brand permanently — the community it serves is too small for a second chance.
AI capabilities involved
Meeting transcript summarization (city council, school board, planning commission)
Newsletter formatting and section drafting from reporter copy
SEO headline generation and meta-description writing
Public-records request template generation
Who uses this
- Solo publisher-reporters doing $80K–$300K on Substack or Ghost, personally writing every story and managing the newsletter
- 2–3 person news operations with one editor and one or two beat reporters covering city hall, schools, and local business
- Nonprofit local news organizations with 4–6 staff doing $500K–$1M in subscription + grant revenue
- Freelance journalists running their own paid newsletter while contributing to larger outlets
SaaS alternatives on the market
Real products you can sign up for today — with current 2026 pricing, honest pros and cons.
Substack
Solo publisher-reporters starting from zero who want the Substack discovery network and don't want to manage technical infrastructure
Yes — free to publish; 10% of paid subscription revenue
10% of subscription revenue (no flat monthly fee)
Pros
- +Built-in subscription infrastructure, paywall, and reader discovery network at zero upfront cost
- +Substack's recommendation algorithm actively routes new readers to newsletters in the same category
- +No migration friction for readers — they subscribe within the Substack ecosystem
- +Native podcast and video hosting on paid plans
Cons
- −10% of revenue is expensive at scale — a $500K operation pays $50K/year to Substack
- −Limited customization of newsletter layout and subscriber management compared to Ghost
- −No self-hosted option — if Substack changes terms or pricing, migration is costly
- −Export of subscriber email list is possible but requires manual steps — you don't 'own' the relationship as clearly as on Ghost
Ghost Pro
Established local news operations with 500+ subscribers who want flat-fee economics and full data ownership
No (self-hosted Ghost is free and open-source)
$9/mo (Starter: up to 500 members)
Pros
- +Flat monthly fee regardless of subscriber count — economics improve dramatically as subscriber base grows
- +Full newsletter customization, paywall logic, and membership tiers with more flexibility than Substack
- +You own your subscriber data and relationship — no platform dependency
- +API available for custom integrations (limited but functional)
Cons
- −No built-in discovery network — you have to grow your own audience entirely
- −Self-hosted Ghost (free) requires server management; Ghost Pro handles this but limits customization
- −Zapier integrations for Otter.ai → Ghost newsletter workflow add $20–$49/mo
- −Ghost's AI features are basic — you still need external Claude/ChatGPT for substantive writing assistance
Otter.ai
Local news operations covering 2–8 government meetings per month where transcription is the primary time bottleneck
Yes — 300 minutes/mo transcription, 30 min per conversation
$16.99/mo (Pro) — verified 2026 pricing
Pros
- +Best-in-class accuracy for in-person and Zoom meeting transcription in noisy environments
- +Speaker identification labels each speaker separately — critical for city council meeting summaries
- +Auto-summary generation highlights key decisions and action items
- +Integrates with Zoom and Google Meet for automatic recording
Cons
- −Accuracy degrades with poor audio, heavy accents, or overlapping speakers — always review before using as a reporting source
- −Auto-summary can miss the most important decisions in favor of most-discussed topics — don't skip the full transcript review
- −Pro plan at $16.99/mo caps at 1,200 minutes of transcription — a high-output newsroom covering 8+ meetings/month may hit the cap
- −GDPR/CCPA implications for transcribing public officials — consult your state's recording consent laws
The AI stack
Local news needs exactly 2–3 AI layers: a transcription tool for meeting capture, a text LLM for summary and copy assistance, and the newsletter platform itself. Don't over-engineer a workflow for a 2-person newsroom.
Meeting transcription
Converts government meeting audio into searchable, attributable text that feeds the AI summarization step
Otter.ai Pro
$16.99/mo (1,200 min transcription)Newsrooms covering 4–8 government meetings per month via Zoom or in-person with a phone recording
Whisper (OpenAI, via API)
$0.006/min of audioTechnical reporters comfortable with a basic automation script who cover 20+ hours of meetings per month
Our pick: Otter.ai Pro at $16.99/mo for most local news operations. Switch to Whisper API only if you're covering 20+ meeting hours/month and willing to set up a basic automation — the cost drops from $16.99/mo to under $8.
Editorial AI assistant (summarization, copy, SEO)
Converts transcripts to meeting summaries, formats newsletter sections, generates SEO headlines, drafts public-records templates
Claude Sonnet 4.6
$3.00 / $15.00 per M tokens (input/output)Meeting summaries, public-records request drafts, and any long-form document comprehension task
GPT-5.4 mini
$0.75 / $4.50 per M tokensSEO headlines, meta-descriptions, social promo copy, and subscriber email subject lines
Gemini 3.5 Flash
$1.50 / $9.00 per M tokensExtracting structured data from government documents or dense PDF reports
Our pick: Claude Sonnet 4.6 via Claude Pro ($20/mo) for meeting summaries and long-document work. GPT-5.4 mini via ChatGPT Plus ($20/mo) for SEO and newsletter formatting tasks. Both are needed — they serve different tasks in the editorial workflow.
Reference architecture
The workflow is transcript-in, summary-out, newsletter-formatted: Otter.ai captures the meeting, the reporter uploads the transcript to Claude, Claude generates a structured summary with key votes and quotes flagged for verification, the reporter verifies every attributed statement against the transcript, then the approved summary feeds into the newsletter draft. The hardest discipline is the human verification step — under deadline pressure, reporters are tempted to skip it.
Reporter attends or records the government meeting; Otter.ai Pro auto-transcribes via Zoom or phone recording
Otter.ai Pro (auto-transcription) or phone recording appSpeaker labels are critical — make sure Otter.ai has identified the mayor, council members, and key community members correctly before proceeding. Errors in speaker attribution cascade into AI summary errors.
Reporter uploads transcript to Claude Sonnet 4.6 and runs the meeting summary prompt
Claude Pro browser (1M context handles full transcripts)Prompt specifies: extract key votes (with tally), key decisions, public comment themes, and any notable statements — formatted as a structured summary with the speaker name and transcript timestamp for every attributable claim.
Reporter verifies every attributed quote and vote tally against the original transcript
Reporter review — non-negotiableClaude occasionally misattributes statements to the wrong speaker when Otter.ai speaker labels are ambiguous. Every attribution in the summary must be cross-checked against the timestamp in the original transcript before the summary is published or used as reporting notes.
Approved summary exported to reporter notes; reporter writes the actual story from their notes and the verified summary
Reporter's own writing workflow (Google Docs, Notion, etc.)The AI summary is a reporting aid, not a publishable story. The reporter writes the story. AI cannot substitute for the reporter's editorial judgment on newsworthiness, source relationships, and community context.
ChatGPT Plus generates 5–7 SEO headline variants and a meta-description from the reporter's published story
ChatGPT PlusPrompt includes: the story, target keyword phrase, and the outlet's SEO style guide. Reporter selects the best headline variant and adjusts the meta-description before publishing.
Newsletter section formatted: ChatGPT Plus drafts the newsletter intro paragraph and story teasers from the week's published stories
ChatGPT Plus + Substack or Ghost editorReporter pastes the week's 3–5 story headlines and first paragraphs; ChatGPT generates the newsletter intro and 2-sentence teasers for each story. Reporter edits for voice and accuracy before scheduling.
Subscription retention: ChatGPT drafts subscriber win-back emails for lapsed or pausing subscribers
ChatGPT Plus + Substack or Ghost native emailReporter personalizes the win-back with a recent story that would matter to the subscriber's stated interests. AI handles the structure; the reporter adds the personal hook.
Estimated cost per request
~$0.05–$0.15 per meeting summary via API (Claude Sonnet 4.6: ~15,000 tokens for a 90-minute council meeting transcript + 2,000 token output). At 8 meetings/month, direct API cost is $5–$10/month — subscription plans are cheaper at this volume.
Cost calculator
Drag the sliders to model your actual usage. The numbers update in real time so you can stress-test economics before writing a single line of code.
Models a 1–3 person local news operation covering 4–8 government meetings per month and publishing a weekly or twice-weekly newsletter. Fixed costs are the subscription tools; per-unit costs reflect API usage if you exceed subscription limits.
Estimated monthly cost
$333
≈ $3,991 per year
Calculator notes
- Fixed stack cost on Ghost Pro is $82/mo ($984/yr) — within the $50–$100/mo target
- On Substack at 500 paid subscribers × $5/mo × 10% fee = $250/mo — dramatically more expensive than Ghost Pro at $25/mo above ~250 paid subscribers
- Otter.ai Pro at 1,200 min/mo covers ~20 hours of meetings; Business plan at $30/mo covers 5,000 min for high-coverage newsrooms
- The real cost variable is reporter time — at 4–6 hours saved per week × $40/hr = $640–$960/month in labor value recovered
Build it yourself with vibe-coding tools
You already have a publishing platform. You need two prompt templates and a verification habit — set them up on your next city council meeting and you'll have a working system by the end of the week.
Time to MVP
1–2 evenings of setup
Total cost to MVP
$40 (ChatGPT Plus + Otter.ai free trial, first month)
You'll need
Starter prompt
You are an editorial assistant for [OUTLET NAME], a local news outlet covering [CITY/REGION]. I am going to give you the transcript of a [CITY COUNCIL / SCHOOL BOARD / PLANNING COMMISSION] meeting from [DATE]. Please generate a structured meeting summary with the following sections: 1. KEY VOTES AND DECISIONS (with vote tally if mentioned in transcript) 2. MAJOR AGENDA ITEMS DISCUSSED (2-3 sentences each) 3. PUBLIC COMMENT THEMES (aggregate — do not attribute individual comments without clear speaker identification) 4. NOTABLE STATEMENTS (direct quotes only — include speaker name and the timestamp from the transcript so I can verify each one before I use it) 5. WHAT HAPPENS NEXT (any scheduled follow-ups, votes, or deadlines mentioned) IMPORTANT RULES: - Only use direct quotes you can find verbatim in the transcript. Flag any quote you're uncertain about with [VERIFY]. - Do not attribute statements to a speaker unless their name is clearly identified in the transcript. - Do not add context, background, or interpretation that isn't in the transcript — that is my job as the reporter. - If Otter.ai's speaker labels seem confused (e.g., multiple statements that don't make sense for one speaker), note this as [SPEAKER UNCLEAR] rather than guessing. Here is the transcript: [PASTE OTTER.AI TRANSCRIPT]
Paste this into Claude
Follow-up prompts (run in order)
- 1
SEO and newsletter weekly batch (run after each story is published): 'I have published the following stories this week on [OUTLET NAME]: [PASTE HEADLINE + FIRST PARAGRAPH for each story]. For each story, generate: (1) 3 SEO headline variants optimized for the phrase "[CITY] [TOPIC]" (e.g., "Austin city council" or "Denver school board"), (2) a 155-character meta-description, (3) a 2-sentence newsletter teaser. Format as a table.'
- 2
Public-records request template (run per FOIA/sunshine request): 'I want to file a public-records request in [STATE] for [DESCRIBE RECORDS: e.g., all communications between the city manager and [DEVELOPER] from January 1, 2026 to present]. Please draft a records request letter that: cites the specific [STATE] Open Records / Sunshine / FOIA statute, requests electronic format where possible, asks for an itemized fee estimate before any charges are incurred, and sets a deadline for a response acknowledgment. I will add my contact information and the agency address.'
- 3
Subscriber win-back sequence (run monthly for lapsed subscribers): 'I run a local news newsletter in [CITY] called [OUTLET NAME]. Our recent coverage includes: [PASTE 3-5 RECENT STORY HEADLINES]. Please write 3 email variants for subscribers who paused or lapsed in the last 60 days. Each email should: be 100 words or less, open with a specific recent story that matters to a local reader, remind them what we cover and why it matters, and include a CTA to reactivate. Do not use generic "we miss you" openers.'
Expected output
A repeatable weekly workflow that recovers 4–6 hours of reporter time from transcription and newsletter formatting, and produces verifiable meeting summaries, SEO-optimized headlines, and subscriber retention emails — without the reporter writing any of the operational copy from scratch.
Known gotchas
- !AI meeting summaries will occasionally misattribute quotes to the wrong speaker when Otter.ai speaker labels are ambiguous — every attribution requires a timestamp cross-check before use
- !AP, Reuters, and the Institute for Nonprofit News (INN) all require AI-use disclosure in published content — check your association's guidelines and update your editorial policy before deploying AI in your workflow
- !Never let AI write the actual story — meeting summaries are reporting aids, not publishable content. One fabricated quote ends a local news brand permanently
- !Substack's 10% subscription fee becomes financially painful above 250 paid subscribers at $5/mo — plan your Ghost migration before you need it
- !Claude Pro and ChatGPT Plus usage limits both reset every 5 hours — high-output news weeks (big council vote, breaking story) will exhaust the allowance; budget $10–$20 in API credits as overflow
- !Otter.ai transcription accuracy degrades significantly with poor meeting audio — external microphone or phone placed near the podium dramatically improves output quality, and therefore summary quality
Compliance & risk reality check
Local news AI compliance is about journalistic integrity, AI-use disclosure to readers, and subscription billing law — not FDA or financial regulation. The risks are existential for a news brand if the disclosure and verification disciplines fail.
AI-use disclosure to readers and industry standards
The Associated Press, Reuters, Institute for Nonprofit News (INN), and Local Media Association all have published guidelines requiring disclosure when AI is used in published content. Industry standard in 2026 is an editorial note when AI was used in the reporting or writing process — not just in headline generation. A local news outlet that does not disclose AI use and is exposed by a competitor or reader will suffer reputational damage disproportionate to its size because it serves a community that knows the people being covered.
Mitigation: Adopt a written editorial AI policy that specifies: what AI can be used for (transcription assist, summary drafts, SEO copy, newsletter formatting), what it cannot be used for (writing stories, generating quotes, creating sourced content), and how AI use will be disclosed to readers. Add a standing disclosure line to your newsletter and website footer. Review INN's AI use policy template as a starting point.
Source verification and libel risk from AI-generated content
An AI-generated meeting summary that misattributes a statement to a public official — even a factual-sounding error — is potential defamation if it damages the official's reputation and is published without verification. Local news outlets are not protected by the same institutional legal resources as large media companies; a single defamation claim can bankrupt a hyperlocal operation. AI hallucinations in Otter.ai transcripts + Claude summarization create a compounding error risk.
Mitigation: Establish and enforce the verification rule: no AI-generated attribution goes into published content without a reporter cross-checking the exact quote against the original transcript or recording. Add [VERIFY] flags to your summary prompt as a structural reminder. Document your verification process in your editorial policy.
Subscription auto-renewal disclosure (ROSCA and California ARL)
Both Substack and Ghost offer automatic subscription renewal, which subjects local news publishers to ROSCA (Restore Online Shoppers' Confidence Act) at the federal level and the California Automatic Renewal Law (ARL) at the state level — even if you're not based in California, if any subscribers are California residents, California ARL applies. Required disclosures include clear terms before checkout, easy cancellation, and specific notice before a free trial converts to a paid subscription.
Mitigation: Substack handles the subscription mechanics and ARL compliance on your behalf as part of their 10% fee — this is one of the underappreciated values of Substack for small news operators. Ghost Pro handles this as well. If you use a custom payment system or subscription manager, you need a lawyer to review your ARL disclosures and cancellation flow.
Reader data privacy (CCPA, state laws)
Your subscriber email list is personal data subject to CCPA if any subscribers are California residents. Substack and Ghost both have CCPA-compliant data processing. The additional consideration in 2026 is that pasting subscriber data into AI prompts (e.g., giving Claude a subscriber's name and lapsed date for a win-back email) creates a data minimization obligation.
Mitigation: Do not paste subscriber PII (names, email addresses) into ChatGPT or Claude prompts — use anonymized identifiers or first-name-only references. The actual email send happens through Substack or Ghost, not through the AI tool. Review Anthropic's and OpenAI's data processing terms if GDPR applies to your EU subscribers.
State recording consent laws for meeting transcription
Recording a government meeting for Otter.ai transcription is generally lawful under state open-meetings laws, but the rules vary by state. Some states require all-party consent for recording (12 states as of 2026), while others are one-party consent. Recording a public official in a public meeting is generally permitted, but recording private citizen testimony at a public meeting requires more care. Otter.ai's terms do not substitute for your state's recording law.
Mitigation: Confirm your state's recording consent law before using Otter.ai in public meetings. Announce your recording in open public meetings as a best practice. If in doubt, rely on public meeting minutes (which are public records) rather than your own recording for attributable quotes.
Build vs buy: the real math
6–10 weeks
Custom build time
$13,000–$25,000
One-time investment
24–60 months
Breakeven vs buying
The SaaS stack on Ghost Pro costs $82/mo ($984/year). A custom build at $13K–$25K requires 13–25 years of SaaS subscriptions to break even on pure tool cost — clearly the wrong math for almost all local news operations. The real case for a custom build is at $500K+ subscription revenue where Ghost/Substack's native paywall logic is genuinely insufficient: tiered institutional subscriptions, public-records request tracking integrated into the CMS, and AI-use disclosure logging baked into the publishing workflow. At that scale, a $13K–$25K build recovers its cost in 6–18 months through reduced platform fees (Ghost Pro at $50/mo vs Substack's $50K/year cut on $500K revenue) and operational automation. Note that journalism-specific foundation grants (INN, Knight Foundation, MacArthur) increasingly fund technology infrastructure for local news — the right first step is a grant application, not an out-of-pocket build.
Skip the DIY — RapidDev builds the production version
A Lovable MVP gets you a demo. Production needs auth that doesn't leak data, AI calls that don't bankrupt you, observability when models drift, and code you can audit. That's what we ship.
Discovery call (free)
30 minWe map your exact Local News Website use case: who uses it, target volume, AI model choice, integrations, compliance scope. You get a detailed scope document and fixed-price quote within 48 hours.
AI-accelerated build
6–10 weeksOur engineers use Claude Code, Lovable, and custom tooling to ship 3–5x faster than agencies. You see weekly progress in a staging environment — not a black box.
Launch + handoff
1 weekWe deploy to your infrastructure, transfer the GitHub repo, set up CI/CD and monitoring, and train your team. You own 100% of the source code, prompts, and model configurations.
What you get
Timeline
6–10 weeks
Investment
$13,000–$25,000
vs SaaS
ROI in 24–60 months
30-min call. Fixed-price quote within 48 hours. No commitment.
Frequently asked questions
How much does it cost to set up AI tools for a local news website?
The full stack — Ghost Pro ($25/mo) + Otter.ai Pro ($16.99/mo) + ChatGPT Plus ($20/mo) + Claude Pro ($20/mo) — runs $82/mo or $984/year. On Substack, the platform fee replaces Ghost Pro but scales with revenue (10% of subscription income). A custom CMS with AI integration costs $13K–$25K upfront and only makes sense above $500K subscription revenue.
How long does it take to set up the AI editorial workflow?
One evening to create accounts and build your 3 core prompt templates (meeting summary, SEO batch, win-back email). One city council meeting to test and refine the transcript-to-summary workflow. Plan to iterate your prompts for 2–3 meeting cycles before they're reliable enough for production use. A RapidDev custom build takes 6–10 weeks.
Can RapidDev build a custom CMS with AI integration for my local news operation?
Yes — RapidDev has shipped 600+ applications and can build a custom CMS with Claude API integration, meeting-transcript automation, AI-use disclosure logging, and tiered paywall management. The standard build is $13K–$25K with a 6–10 week timeline. Book a free 30-minute consultation at rapidevelopers.com to scope your specific coverage model and subscriber volume.
Can AI write stories for my local news website?
No — and this is the hardest line in local news AI use. AI can summarize a meeting transcript, draft newsletter copy, generate SEO headlines, and write public-records request templates. It cannot write the actual reporting, generate quotes attributed to sources, or substitute for a reporter's editorial judgment about newsworthiness. One AI-fabricated quote in a published story will end a local news brand permanently in the community it serves.
Do I need to disclose AI use to my readers?
Yes — the AP, Reuters, INN, and Local Media Association all require disclosure when AI is used in published content. In 2026, the industry standard is an editorial note when AI contributed to the reporting or writing process. This is not just an ethical obligation; it's increasingly a requirement for membership in journalist professional associations and for eligibility for foundation grants. Adopt a written editorial AI policy before deploying AI in your workflow.
Should I be on Substack or Ghost for a local news operation?
Start on Substack for the discovery network and zero-friction subscription setup. Migrate to Ghost Pro when your paid subscriber revenue exceeds roughly $3,000/month — at that point, Substack's 10% fee ($300+/mo) is more expensive than Ghost Pro ($25–$50/mo) plus the minor friction of managing your own subscriber relationship. Substack's 10% is a worthwhile cost at 50–200 paid subscribers; it becomes a significant drag above 300.
How accurate is Otter.ai for government meeting transcription?
Otter.ai Pro is accurate enough for structured meeting environments (council chambers, Zoom calls with a good mic) — expect 90–95% word accuracy in good conditions. Speaker attribution is the weaker link: Otter.ai struggles when multiple people speak over each other or when council members don't identify themselves before speaking. Every AI-generated meeting summary must be cross-checked against the original transcript before any attribution is published. Never rely on Otter.ai's auto-summary as your sole source of record.
Want the production version?
- Delivered in 6–10 weeks
- You own 100% of the code
- AI cost monitoring built in
30-min call. No commitment.