# The Intelar Editorial Playbook

> The operating system for every story we publish. Read this once. Reference it always.

This playbook defines what a Intelar story is — and what it isn't. It covers voice, structure, images, sourcing, and the quality gates a piece must pass before it ships. Nothing here is suggestion; everything is standard.

Sections:

1. [The Voice](#1-the-voice)
2. [Anatomy of an Intelar Story](#2-anatomy-of-an-intelar-story)
3. [The Image System](#3-the-image-system)
4. [Free Stock Image Sources](#4-free-stock-image-sources)
5. [The 30-Second Image Selection Test](#5-the-30-second-image-selection-test)
6. [Image Processing Pipeline](#6-image-processing-pipeline)
7. [Pre-Publish Quality Gate](#7-pre-publish-quality-gate)
8. [Common Failure Modes](#8-common-failure-modes)

---

## 1. The Voice

Intelar has one voice. Every contributor adapts to it.

### Three principles

**1. Declarative, not hedged.**
We write *"Stripe is rebuilding its stack on Claude"* — not *"Stripe appears to potentially be considering a rewrite."* Confidence is earned by reporting, not avoided by qualifier-stacking. If we can't say it cleanly, we don't say it yet.

**2. Named over abstract.**
We name companies, people, products, dollar amounts, dates. *"Patrick Collison hired three Anthropic engineers in Q1 2026"* beats *"the company has been investing in AI talent."* The specificity is the value.

**3. Active over passive.**
We say who did what. *"Anthropic shipped Skills"* — not *"Skills was launched by Anthropic."* The agent goes first.

### What we never do

- **Hype verbs.** No *crushes*, *explodes*, *destroys*, *revolutionizes*, *disrupts*.
- **Exclamation marks.** Ever.
- **Emoji in prose.** Reserved for UI labels, never editorial body.
- **"In today's fast-paced world."** Or any phrase that survives without editing.
- **"We at Intelar believe..."** Editorial voice is implicit. Authorial *we* only when reporting on Intelar methodology.
- **Listicle headlines.** *"10 ways to..."* belongs on Buzzfeed. We write *"How operators ship faster."*
- **Clickbait verbs in headlines.** *"You won't believe"*, *"This one trick"*, *"What happened next."* Auto-reject.

### Sentence rhythm

Vary length. Short punches alongside longer reflective lines. The pattern below works:

> Anthropic shipped Skills three weeks ago. By Tuesday morning, sixty per cent of LangChain deployments inside the Fortune 500 buyers we surveyed had migrated. The orchestration layer, it turns out, was a transit station — not a destination.

Three sentences. Eleven words, twenty-six words, eighteen words. Rhythm.

Paragraphs run two to four sentences. One idea per paragraph. The reader should be able to skim H2 → first line of paragraph → understand the argument.

### Numbers, dates, names

- **Dates**: spelled out in body (*"20 May 2026"*), abbreviated only in metadata (*"5/20/26"*).
- **Numbers**: spell out one through nine; numerals for ten and above. Currency in full notation (*$13.2B*, *$4.2M*, not *"thirteen billion"* in prose).
- **First mention**: full name + role + company. *"Patrick Collison, CEO of Stripe"*. Subsequent: surname only — *"Collison"*.
- **Companies**: capital case as the company brands itself. *"OpenAI"*, *"DeepMind"*, *"xAI"* — match the source.

### When in doubt

Read it back aloud. If it sounds like Forbes, Bloomberg, the FT, or *The Atlantic*, it's right. If it sounds like a SaaS landing page or a LinkedIn post, rewrite.

---

## 2. Anatomy of an Intelar Story

Every story has the same skeleton. The skeleton is non-negotiable. The flesh is editorial.

```
1. KICKER          (eyebrow, uppercase, red — category + format)
2. HEADLINE        (Newsreader serif, max 14 words, italic emphasis allowed)
3. DECK            (italic serif, supporting context, max 36 words)
4. HERO IMAGE      (16:11, with caption + photographer credit)
5. BYLINE + META   (author, role, read time, dated, updated stamp)
6. TL;DR BOX       (3-5 bullets — GEO/AI Overview optimization)
7. BODY            (sectioned with H2, drop cap on lede, inline images)
8. PULL QUOTES     (1-3 per piece, used for inflection points only)
9. AFFILIATE CTA   (max 1, after editorial verdict is established)
10. FAQ            (3-5 Q&As, FAQPage schema-wired)
11. AUTHOR BLOCK   (name, role, bio, credentials, citation count)
12. RELATED        (3 internal links — same category / adjacent topics)
```

### Headline rules

- **Max 14 words.** If it doesn't fit, the headline isn't sharp enough yet.
- **Lead with the news, not the angle.** Headline says *what happened*; deck supplies *why it matters*.
- **Italic emphasis** for one or two words that carry the editorial argument. *"The $13B agent economy nobody is **reporting** on."*
- **No question marks** in news headlines. Questions are deflective. Reserve for opinion/explainer pieces only.
- **Avoid colons** in news headlines — they signal a SEO-stuffed listicle. Colons are fine in analysis/dossier headlines where they serve as a literal subtitle separator.

### Deck rules

- The deck is the second pull — it sells the reader on staying.
- One thought, max two sentences, max 36 words.
- Should answer *"why should I care?"* — not summarize the article.
- Italic Newsreader serif. No exceptions.

### TL;DR box

The single highest-leverage GEO move on the page. Optimized for AI Overviews extraction.

- Three to five bullets, each declarative, each standalone.
- Lead bullet states the news.
- Subsequent bullets supply: scale (numbers), actors (named entities), mechanism (how), implication (so-what).
- Each bullet starts with an arrow `→` (rendered by the component, not typed).
- Maximum 22 words per bullet — LLMs truncate longer.

### Body structure

- **Lede paragraph**: drop cap (first letter set in display serif, 4.5rem). Lede must establish the news in under 80 words.
- **H2 sections**: italic emphasis welcome (*"What actually **changed**"*). Section headers should be five to seven words. Each H2 anchors a TOC entry.
- **H3 sub-sections**: optional, used only for analysis pieces with three or more sub-arguments per H2.
- **First mention of every tool, company, or person**: hyperlinked to its `/topic/[slug]` page.
- **Inline images**: every 600–800 words minimum. See §3 for placement rules.
- **Pull quotes**: 1–3 per piece. Pull a sentence that *changes the argument's direction* — not the opener, not the conclusion. Pull-quotes are inflection points.
- **Block quotes (interview)**: attribute with `<cite>` — *— Patrick Collison, Stripe Sessions 2026*.

### Section pacing

| Story type | Word count | Sections (H2) | Inline images | Pull quotes |
| ---------- | ---------- | ------------- | ------------- | ----------- |
| Briefing   | 400–800    | 0–2           | 1             | 0           |
| Field Notes | 1,000–1,800 | 3–4         | 2             | 1           |
| Analysis   | 1,800–3,000 | 5–7         | 3             | 1–2         |
| Dossier    | 3,000–6,000 | 7–10        | 4–6           | 2–3         |
| Review     | 1,500–2,500 | 4–6 + scorecard | 3        | 1           |
| Comparison | 2,000–3,500 | varies + matrix | 2          | 0           |

### Affiliate CTA placement

If the piece contains an affiliate, the CTA goes **after** the section that establishes editorial verdict — never above the fold, never adjacent to the lede.

The CTA itself contains: editor's-pick label, product name, one-line description, button. See the `affiliate-cta` component in [article.html](article.html).

---

## 3. The Image System

Images are not decoration. They are evidence and rhythm.

### Why every story needs imagery

1. **Reader retention** — pieces with imagery every 600–800 words hold attention 38% longer (industry benchmark; we measure).
2. **Editorial trust** — a piece without a hero image reads like a draft. Premium readers register the absence within two seconds.
3. **GEO performance** — Google Images is the second-largest entry surface after search; AI Overviews increasingly include featured imagery from cited sources.
4. **Social distribution** — open-graph and Twitter cards demand a hero. Stories without one get half the click-through.

### Aspect ratios — the only ratios we use

| Slot | Ratio | Min pixel width | Use case |
| ---- | ----- | --------------- | -------- |
| Hero (article header) | 16:11 | 2400px | Every story |
| Inline wide | 16:11 | 1800px | Section break, evidence shot |
| Inline portrait | 4:5 | 1200px | Person, single subject, vertical scene |
| Inline data | 16:9 | 1800px | Charts, data visualization |
| Author headshot | 1:1 | 400px | Byline + author block |
| Card thumbnail | 4:3 | 800px | Related cards, sidebar items |
| Hero ultrawide | 21:9 | 2800px | Dossier covers, longform anchors |

Anything outside these ratios gets re-cropped. The grid is not negotiable.

### Image required at these positions

| Position | Required? | Notes |
| -------- | --------- | ----- |
| Hero (article header) | Always | 16:11 |
| After first H2 | If story > 1,200 words | 16:11 or 4:5 |
| Mid-body (every 600–800 words after) | Yes | Alternate 16:11 and 4:5 |
| Inside affiliate CTA | The product/company glyph | 1:1 |
| Author block | The author headshot | 1:1 |
| Related cards | One per card | 4:3 |

A 2,500-word analysis piece needs: 1 hero + 3 inline + 1 author = **5 images minimum**.

### The Intelar image aesthetic

Photos must look like they would run in:

- *The Atlantic* — moody, restrained, intentional
- *Bloomberg Businessweek* — editorial illustration, conceptual
- *FT Weekend* — natural light, architectural, human-scale
- *Wired* — high-contrast, considered, sometimes abstract

Photos must **not** look like:

- Generic SaaS stock — *team huddled around laptop pointing at screen*
- Robot-with-glowing-brain AI cliché
- *Diverse group of professionals laughing at conference table*
- *Hand holding phone with abstract data overlay*
- Anything on the first page of a "business" Shutterstock search

When the perfect photo doesn't exist, we prefer **abstract / architectural / texture** over a forced literal shot. A close-up of brushed steel reads more credibly than a stock CEO portrait.

---

## 4. Free Stock Image Sources

Every source below has been vetted for license, quality average, and editorial fit. Sources are tiered by reliability and average quality.

### Tier 1 — primary (search here first)

| Source | URL | License | Best for |
| ------ | --- | ------- | -------- |
| **Unsplash** | [unsplash.com](https://unsplash.com) | Unsplash License — commercial OK, no attribution required | The default. 5M+ photos, strongest curation, premium aesthetic. ~70% of our hero images come from here. |
| **Pexels** | [pexels.com](https://pexels.com) | Pexels License — commercial OK, no attribution required | Strong second source. Better business/product/people coverage than Unsplash. |
| **Wikimedia Commons** | [commons.wikimedia.org](https://commons.wikimedia.org) | Varies — must check each (CC0, CC-BY, public domain most common) | Historical, scientific, geographical, news-event photography. Indispensable for context shots. |

### Tier 2 — secondary (when Tier 1 doesn't deliver)

| Source | URL | License | Best for |
| ------ | --- | ------- | -------- |
| **Pixabay** | [pixabay.com](https://pixabay.com) | Pixabay Content License — commercial OK | Huge volume, lower curation. Use when you need a specific subject Tier 1 missed. |
| **Lummi** | [lummi.ai](https://lummi.ai) | Free for commercial use with attribution to source | AI-generated. Useful for tech/abstract concepts where photography fails. |
| **Burst (Shopify)** | [burst.shopify.com](https://burst.shopify.com) | CC0 (public domain) | Business, product, retail. Stronger e-commerce vibe. |
| **Reshot** | [reshot.com](https://reshot.com) | Reshot License — commercial OK | More artistic. Smaller library, higher individual quality. |
| **StockSnap** | [stocksnap.io](https://stocksnap.io) | CC0 | Curated subset of free photography. |

### Tier 3 — specialist (when the story demands a specific archive)

| Source | URL | License | Best for |
| ------ | --- | ------- | -------- |
| **NASA Image Library** | [images.nasa.gov](https://images.nasa.gov) | Public domain | Space, science, satellite imagery, scientific concepts. |
| **Library of Congress** | [loc.gov/free-to-use](https://www.loc.gov/free-to-use/) | Public domain (most) | Historical photography, archival news, 20th-century context. |
| **Smithsonian Open Access** | [si.edu/openaccess](https://www.si.edu/openaccess) | CC0 / public domain | 4M+ items: scientific specimens, historical artifacts, art. |
| **NYPL Digital Collections** | [digitalcollections.nypl.org](https://digitalcollections.nypl.org) | Mostly public domain | Maps, prints, historical photographs. |
| **The Met Open Access** | [metmuseum.org/art/collection](https://www.metmuseum.org/art/collection/search) | CC0 (filter for "Open Access") | Fine art for luxury/culture stories. |
| **Internet Archive Images** | [archive.org/details/image](https://archive.org/details/image) | Varies — most public domain | Vintage photographs, ephemera. |
| **Public Domain Review** | [publicdomainreview.org](https://publicdomainreview.org) | Public domain | Curated rare imagery, often perfect for analysis pieces. |
| **Nappy.co** | [nappy.co](https://nappy.co) | CC0 | Black and brown people — fills a notable gap in mainstream stock libraries. |

### Illustration sources (for abstract / concept stories)

| Source | URL | License | Best for |
| ------ | --- | ------- | -------- |
| **unDraw** | [undraw.co](https://undraw.co) | unDraw License — commercial OK, recoloring built-in | Editable SVGs in any brand color. Best for explainer pieces. |
| **DrawKit** | [drawkit.com](https://drawkit.com) | Mixed — free and paid; check each | Polished editorial illustrations. |
| **Open Peeps** | [openpeeps.com](https://www.openpeeps.com) | CC0 | Hand-drawn people, mix-and-match. |
| **Humaaans** | [humaaans.com](https://www.humaaans.com) | CC0 | Mix-and-match human illustrations. |
| **Storyset** | [storyset.com](https://storyset.com) | Free with attribution | Animated and static editorial illustrations. |

### Icon sources

| Source | URL | License |
| ------ | --- | ------- |
| **Lucide** | [lucide.dev](https://lucide.dev) | ISC — fully free |
| **Tabler Icons** | [tabler.io/icons](https://tabler.io/icons) | MIT |
| **Phosphor** | [phosphoricons.com](https://phosphoricons.com) | MIT |
| **Heroicons** | [heroicons.com](https://heroicons.com) | MIT |

### License quick-reference

- **CC0 / Public Domain**: Use without restriction.
- **Unsplash License**: Commercial use allowed. Do not redistribute the photo *as* stock (i.e., we can't upload it to our own stock library).
- **Pexels License**: Same as Unsplash in practice.
- **CC-BY (Creative Commons Attribution)**: Attribution required. We attribute in the photo credit line.
- **CC-BY-SA (Share-Alike)**: Attribution required + derivatives must be CC-BY-SA. Avoid for our use — creates licensing complexity.
- **CC-BY-NC (Non-Commercial)**: Do not use. Our use is commercial.

When in doubt: **check the file's license page directly**, not the platform's general TOS.

---

## 5. The 30-Second Image Selection Test

This is the checklist that gets run on every candidate image. If a candidate fails any criterion, it's rejected. No appeals.

### Criterion A — Editorial fit (5 seconds)

- [ ] Does it support the article's argument, or merely decorate around it?
- [ ] Would the article suffer without it? (If "no" — it's filler. Reject.)
- [ ] Does it telegraph the story's emotional register? (Analysis is sober; field-notes is curious; review is exact.)

### Criterion B — Aesthetic quality (5 seconds)

- [ ] Composition has either a clear focal point or intentional negative space.
- [ ] Light is natural and intentional (not flash-blasted, not over-HDR).
- [ ] Color palette is harmonious — does not fight the Intelar newsprint background or the editorial red accent.
- [ ] Subject is captured *interestingly*, not literally. (A photo of an office is rarely the right photo for a story about an office.)

### Criterion C — Technical (5 seconds)

- [ ] Minimum width matches the slot (see §3 ratio table — hero needs 2400px).
- [ ] No watermarks, source-platform logos, or vendor stamps.
- [ ] Sharp at the intended display size; no compression artifacts or visible noise.
- [ ] Crops cleanly to the required aspect ratio without losing the subject.

### Criterion D — Distinctiveness (5 seconds)

- [ ] Run a 5-second reverse image search ([Google Lens](https://lens.google.com) or [TinEye](https://tineye.com)). If it appears on the first page of any AI/business publication in the last 12 months — reject.
- [ ] It is not the photo every other publication uses for this topic. (Robot hand touching human hand → auto-reject.)
- [ ] If the source is Unsplash/Pexels — check download count. Anything over 100,000 downloads is too overexposed.

### Criterion E — Ethical / legal (5 seconds)

- [ ] License is verified at the file level (not just the platform).
- [ ] No identifiable people are featured unless they are public figures *in their public role* (a senator at a podium, OK; a private person at a café, no).
- [ ] No copyrighted logos, brand marks, or trademarks visible (no Coca-Cola cans, no Apple logos, no recognizable luxury labels — unless the story is specifically about that brand and we're using the official press image).
- [ ] If the photo includes a recognizable person and the story is sensitive (criminal, medical, financial distress) — additional legal review.

### Criterion F — Intelar brand fit (5 seconds)

- [ ] If displayed against `#FAFAF7` (our newsprint background) with a 1px `#D5D5D2` border — does it feel premium or does it feel like a SaaS landing page?
- [ ] Would *The Atlantic* run this? Would *Bloomberg Businessweek*? If the answer to both is no, reject.
- [ ] Does it survive being printed in grayscale? (Our test for compositional strength.)

### The single-sentence summary test

After running all six criteria, write one sentence that begins: *"This image earns its place because…"*

If you can't finish the sentence in under 15 words, the image isn't strong enough. Reject.

---

## 6. Image Processing Pipeline

Every selected image runs through the same pipeline before publish. No exceptions.

### Step 1 — Crop to ratio

Crop to the exact ratio for the slot. Use the rule of thirds to position the subject. Where in doubt, prefer left-third subject placement (Western reading patterns reward this).

Tools: [Figma](https://figma.com), [Photopea](https://photopea.com) (free, browser-based, Photoshop-compatible), Apple Preview for simple crops.

### Step 2 — Color & light treatment

Intelar applies a **subtle, consistent** treatment to every photo so the page feels cohesive:

```
Brightness:   +2 to +5
Contrast:     +6 to +10
Saturation:   −8 to −12          // mute slightly toward editorial
Warmth:       +3 to +6            // very subtle warmth toward newsprint
Highlights:   −10                 // recover blown highlights
Shadows:      +8                  // open up shadow detail
```

These aren't artistic choices — they're standardization. Every Intelar image looks like it belongs in the same publication.

For photos that resist this treatment (already too warm, too contrasty), apply the inverse to bring them into the standard.

### Step 3 — Add subtle grain (optional but encouraged)

A 2–4% noise overlay across the image. This unifies all imagery and prevents the "too-clean" SaaS look. In Photopea: Filter → Noise → Add Noise, 2-4%, monochromatic, Gaussian.

### Step 4 — Export

| Format priority | Use case | Quality |
| --------------- | -------- | ------- |
| **AVIF** | Primary | 70–80% |
| **WebP** | Fallback | 80–85% |
| **JPEG** | Universal fallback | 82% |

Resolution: export at exactly 2× the intended display size. Hero displays at 1200px wide → export at 2400px. The CDN will serve smaller sizes on smaller viewports via `srcset`.

Filenames: `kebab-case-descriptive-slug.avif`. *Never* `IMG_0234.jpg`. The filename is read by search engines and screen readers.

### Step 5 — Write the caption and alt text

These are different. Both are required.

**Caption** — visible text under the image. Describes what the image is + credit. One sentence, declarative.

> *Patrick Collison at Stripe Sessions 2026, where he announced the Atlas-Agent payments API. Photo · Stripe Press.*

**Alt text** — invisible to sighted users, read by screen readers and indexed by search. Describes the image *as content*, not as decoration. One sentence, no "image of" or "photo of" preamble.

> *Patrick Collison gesturing at a podium against a deep blue stage backdrop, holding a remote presenter.*

**Photo credit** — always included. Format:

- *Photo · Unsplash / Photographer Name*
- *Photo · NASA*
- *Photo · Stripe Press* (when using a company press image)
- *Illustration · Intelar* (when we commission)

### Step 6 — Wire the schema

Every image in an article body gets attached to the article's `ImageObject` schema:

```json
{
  "@type": "ImageObject",
  "url": "https://intelar.news/img/agent-economy-hero.avif",
  "width": 2400,
  "height": 1650,
  "caption": "Patrick Collison at Stripe Sessions 2026...",
  "creditText": "Stripe Press",
  "copyrightNotice": "© Stripe, Inc. Used under fair use."
}
```

The hero image is added to the article's main `NewsArticle` schema as the `image` property.

---

## 7. Pre-Publish Quality Gate

Every story passes through this gate before it goes live. The editor running the gate is not the writer.

### Voice & structure

- [ ] Headline ≤ 14 words; lead with the news.
- [ ] Deck ≤ 36 words; sells the *why*.
- [ ] Lede paragraph establishes the news in under 80 words.
- [ ] No hype verbs (*crushes*, *explodes*, *revolutionizes*).
- [ ] No exclamation marks. No emoji.
- [ ] Active voice throughout.
- [ ] Dates spelled out; currency in full notation.
- [ ] First mention of every entity is fully named + linked.

### Story scaffolding

- [ ] Kicker (red, uppercase, accurate category + format).
- [ ] Byline with author headshot + role + read time + dated stamp.
- [ ] TL;DR box: 3–5 declarative bullets, lead bullet is the news.
- [ ] H2 sections: italic emphasis used sparingly, scroll-margin set for TOC.
- [ ] Drop cap rendered on first paragraph.
- [ ] At least one pull quote per 1,500 words.
- [ ] FAQ section: 3–5 Q&As, wired to `FAQPage` schema.
- [ ] Author authority block at the foot.
- [ ] 3 related links at the bottom, all in-domain.

### Imagery

- [ ] Hero image present, 16:11, ≥ 2400px wide.
- [ ] Inline images every 600–800 words.
- [ ] All images passed the §5 30-second test.
- [ ] All images processed through the §6 pipeline (color + grain).
- [ ] Captions present, credits present, alt text present and substantive.
- [ ] Image filenames are descriptive-kebab-case, not `IMG_0234`.
- [ ] Hero image attached to `NewsArticle` schema as `image`.

### SEO + GEO

- [ ] Page `<title>`: headline + ` — Intelar`, ≤ 60 characters.
- [ ] Meta description: deck condensed, ≤ 155 characters.
- [ ] Canonical URL set.
- [ ] OG image (1200×630, generated from headline template) present.
- [ ] `NewsArticle` schema fully populated (datePublished, dateModified, author, publisher, wordCount).
- [ ] `BreadcrumbList` schema present.
- [ ] `FAQPage` schema present if piece contains FAQ.
- [ ] Internal links: minimum 3, maximum 8, all relevant, all to existing pages.
- [ ] External links: open in same tab by default (no `target="_blank"` unless to a tool/affiliate). Affiliate links go through `/go/[slug]`.

### Affiliate

- [ ] Affiliate CTAs (if any) placed *after* editorial verdict.
- [ ] No more than one inline affiliate CTA per 1,200 words.
- [ ] Disclosure label visible on every affiliate component.
- [ ] Sitewide affiliate disclosure present in footer.

### Final read

- [ ] One editor reads the piece aloud, top to bottom. Anywhere the reader stumbles, the writer fixes.
- [ ] The piece is checked at 375px viewport (mobile), 768px (tablet), and 1440px (desktop). Imagery and headlines must work at all three.
- [ ] Core Web Vitals: LCP < 1.8s on the article. If a hero image is dragging LCP, optimize before publish.

If every box checks, the piece publishes.

---

## 8. Common Failure Modes

The eight ways a Intelar story most often goes wrong — and how to catch them.

**1. The decorative-image failure.** The hero looks "nice" but doesn't tell us anything. The reader's eye slides off. Fix: ask *"what does this image add that the headline doesn't?"*  If the answer is *"it makes the page look full,"* find a different image.

**2. The literal-image failure.** Story is about agent economy → image is a stock photo of a robot. The image is doing the headline's job, not its own. Fix: prefer architecture, texture, or abstract photography over literal depiction.

**3. The overprocessed-photo failure.** Image has obvious HDR halos, oversaturated colors, or aggressive sharpening. Reads as amateur. Fix: revert to original, apply our §6 standard treatment, stop.

**4. The clichéd-stock failure.** Robot hand touches human hand. Diverse group laughs at conference table. CEO at floor-to-ceiling window. Auto-reject these regardless of how well they fit the topic.

**5. The hedged-headline failure.** *"How AI might be changing how some companies could think about their stack."* No verbs, all conditionals. Fix: rewrite declaratively. *"AI is rewriting the enterprise stack."*

**6. The unfinished-lede failure.** The first paragraph is throat-clearing — *"In today's rapidly evolving landscape..."* Drop the first 80 words and start with paragraph two.

**7. The missing-TLDR failure.** Long piece ships without a TL;DR box. Half the GEO opportunity is forfeited. Hard requirement: every piece > 800 words has a TL;DR. No exceptions.

**8. The thin-FAQ failure.** FAQ section has questions like *"Why is AI important?"* — generic, low-value. Fix: every FAQ question must be one a real reader would type into a search bar after reading the piece. If we wouldn't search it, we don't FAQ it.

---

## Appendix — Quick image search workflow

A repeatable five-minute process for finding the right image.

1. **Brief** (30 seconds). Open a scratch note. Write three things:
   - *Subject*: the literal thing the photo shows.
   - *Mood*: serious / curious / exact / atmospheric.
   - *Slot*: hero / inline / portrait / data.

2. **Search Tier 1** (90 seconds). Open Unsplash → Pexels → Wikimedia in three tabs. Search for the *abstract* concept, not the literal article topic. (*"abandoned warehouse"* over *"AI infrastructure"*.)

3. **Shortlist** (60 seconds). Open 6–10 candidates as bookmarks or screenshots. Stop searching — more candidates rarely improve the decision, and they delay it.

4. **30-second test** (90 seconds). Run §5 on each candidate. Most will fail B (aesthetic) or D (distinctiveness). Survivors usually number 1–2.

5. **Decide and download** (30 seconds). Pick the strongest. Download at the highest available resolution.

6. **Process** (3 minutes). Run §6 pipeline. Export. Name. Caption + alt + credit.

A skilled editor does this in five minutes. A new contributor in fifteen. Either way, the bar is the bar.

---

*Maintained by the Intelar Editorial Standards Council. Version 1.0, May 2026. Revisions tracked in the changelog at* `/docs/playbook-changelog.md`.