A WordPress post without a featured image gets buried in social shares, in news listings, in the related-posts carousels of other plugins. Yet finding (or commissioning) a fresh image for every article is one of the slowest, dullest parts of running a blog. AI Content Generator solves this by hooking DALL-E 3 directly into your campaign: every new post the plugin writes can also generate its own bespoke cover, with a prompt template you control, in the aspect ratio your theme expects, stored in the WordPress media library and assigned as featured_media automatically.
This tutorial walks through the entire feature: where to find the image generation settings, how to write a prompt template that produces on-brand covers, how to control cost and aspect ratio, how to add inline body images, what to do when DALL-E refuses a prompt, and how to budget for image generation so it never surprises you on the OpenAI bill.
The assumption is that you already have an OpenAI API key configured in AI Content Generator Settings — DALL-E 3 is billed under your OpenAI account, exactly like GPT-4o or any other OpenAI model. If you haven’t set the key up yet, do that first; everything below works the moment the OpenAI connection is green.
What DALL-E 3 brings to a WordPress workflow
DALL-E 3 is the latest image-generation model from OpenAI at the time AI Content Generator was released. Compared to its predecessors it understands natural language prompts much more reliably: you can describe a scene in plain English and it composes a coherent image, including legible (mostly) typography and structured layouts. For a WordPress workflow that means three concrete advantages.
First, you can describe the image at a high level (“editorial cover for a tutorial about WordPress automation, minimal vector style, soft gradient, central concept”) and trust the model to fill in the details. Second, the model is multi-aspect: it can output square (1024×1024), portrait (1024×1792) or wide (1792×1024) — the last one matches what most WordPress themes expect for hero images. Third, DALL-E 3 produces consistently styled outputs when fed the same style instructions, which is what makes “every post has a cover that looks like it belongs on the same site” possible at all.
Step 1 — Open the campaign and find the Featured image panel
Inside a campaign editor (either a new campaign or an existing one), scroll the main column until you see the panel titled Featured image + inline images. This is the block that controls every aspect of the cover (and any inline body image) for posts produced by this campaign. The configuration is per-campaign, so different campaigns can have completely different image styles.
Step 2 — Pick DALL-E 3 as the image source
The first dropdown, Image source, has four options: No image (skip cover entirely), Media Library random (pick a random existing media library image, optionally filtered by tag), Stock provider (Unsplash / Pixabay), AI generation. Pick AI generation (DALL-E 3).
As soon as you choose AI generation, the panel reveals the image-AI-provider sidebar metabox: it confirms DALL-E 3 is configured (it inherits the OpenAI key you saved in Settings) and shows the per-image cost so you can budget at a glance. At the time of writing, DALL-E 3 wide HD costs $0.08 per image — a 30-articles-per-month campaign therefore adds $2.40 to your OpenAI bill purely for covers.
Step 3 — Choose the aspect ratio your theme expects
Most modern WordPress themes use a wide hero ratio for the featured image (somewhere around 1.7:1 to 16:9). Pick 1792×1024 (wide) in the aspect-ratio dropdown — that’s the closest DALL-E 3 output to a typical blog hero. If your theme uses square thumbnails on category pages, that’s fine: WordPress generates the square crop automatically from the wide original.
If you publish to a vertical-first surface like Pinterest or Instagram preview, pick 1024×1792 (tall). The plugin remembers your choice per campaign, so a “blog posts” campaign can output wide while a “social-ready” campaign outputs tall — both from the same WordPress site.
Step 4 — Quality: HD or standard?
DALL-E 3 offers two quality levels: Standard and HD. Standard is cheaper ($0.04 per wide image) but the difference shows up at full-bleed sizes: skin texture, fabric, lighting gradients all look softer. HD is twice the price but produces images you can blow up to 2× without obvious quality loss.
The pragmatic rule: if your theme renders the featured image larger than 800px wide, use HD. If your theme only ever shows it at thumbnail size (small blog cards), Standard is fine and you save half the cost.
Step 5 — Write a prompt template that produces on-brand covers
This is the single most important setting in the panel. The prompt template is sent to DALL-E 3 verbatim, with two placeholders filled in per post: {title} is replaced by the article title and {focus_keyword} by the SEO focus keyword if you set one.
A good template anchors three things: the editorial intent, the visual style and the no-go list. Here is a template that produces consistent results for a SEO-oriented blog:
Editorial cover for a blog article titled "{title}".
Style: minimal vector illustration, soft gradient background,
central concept symbol, clean typography hint, brand palette.
No real human faces. No brand logos. No watermarks.
Aspect: wide, optimised for blog post hero, balanced negative space on the right.
The “no faces / no logos / no watermarks” line matters: DALL-E refuses prompts that ask for celebrity faces or branded products, and saying it explicitly avoids accidental refusals when an article title mentions a person or a brand. The “balanced negative space” hint helps when your theme overlays the article title on top of the image — leaving the right side cleaner gives the typography room to breathe.
Step 6 — Use style presets to enforce consistency across campaigns
The Style preset dropdown gives you reusable building blocks: Editorial illustration, brand colours, Photorealistic, natural light, Cinematic, dramatic lighting, Flat 3D, isometric, Custom. When you pick a preset, the plugin prepends a short style descriptor to your prompt template, so you can keep your template focused on the editorial concept and let the preset handle “what the image should look like”.
For a multi-campaign site, the convention is: one preset per campaign, never mix. That way blog readers immediately recognise which content cluster they’re on — “all the tutorial articles use the editorial illustration style; all the news articles use the cinematic style; all the comparison articles use the flat 3D style”. Recognisability beats variety in editorial design.
Step 7 — Add inline images for long-form posts (optional)
For pillar articles longer than 1,500 words, a single hero image isn’t enough. Set the Inline images count to 2 or 3 and the plugin asks DALL-E to also generate that many supporting images, evenly distributed across the article H2 sections. Each inline image gets its own auto-derived prompt based on the H2 title and the surrounding paragraph, so the visuals stay relevant to what the reader is currently reading.
Be mindful of cost when enabling this. Three inline HD images add $0.24 per article on top of the cover — for a 30-articles-per-month campaign that’s an extra $7.20. Sometimes worth it for high-converting cornerstone content, almost never worth it for short news roundups.
Step 8 — Run a test generation and review the output
Save the campaign, then click Generate content for selected titles with at least one test title queued. The plugin generates the text first, then runs the image generation pass: each post gets its DALL-E cover saved as a new media library item and assigned as featured image automatically.
Open the draft post in Posts → Drafts and inspect the cover. If it looks off-brand, tweak the prompt template — small changes have big effects. Common adjustments: add the brand primary colour by name (“teal accents”), specify a composition (“subject left-aligned, right side cleaner”), or describe what NOT to include (“no text in the image, no UI mockups”).
Troubleshooting common errors
- content_policy_violation — DALL-E refused the prompt. Usually because the article title mentions a public figure, a competitor brand, weapons, or anything in OpenAI’s banned categories. Add anti-instructions in the prompt template (“interpret titles abstractly, do not depict real persons”) or fall back to Media Library random for that specific campaign.
- rate_limit_error on images — DALL-E 3 has stricter rate limits than text models. Raise the Interval between generations in the campaign settings to spread out requests over more time.
- The image doesn’t match the article topic — your prompt template is too generic. Add
{focus_keyword}to the prompt and make sure each article has a focus keyword set. - The featured image is missing entirely — check the WordPress media library: if the image is there but not attached, the campaign may have hit a permission error setting
featured_media. Run a manual Reassign featured from the Logs panel.
Keeping image generation costs predictable
- Default to Standard quality for high-volume short-form. News roundups, listicles, FAQ posts rarely need HD; saving half the image cost on the long tail adds up fast.
- Reserve HD + inline images for cornerstone content. Pillar articles you’ll link to for years deserve the budget; a five-paragraph news brief doesn’t.
- Cap monthly image spend at the OpenAI level. The OpenAI billing dashboard lets you set a monthly hard limit that covers both text and image usage. A misconfigured campaign that hammers DALL-E in a loop gets stopped at the wallet level.
You’re done
That’s the full flow: pick DALL-E 3 as the image source, choose the aspect ratio and quality, write a prompt template that uses {title}, lock in a style preset, optionally add inline images, run a test campaign and adjust the template based on the first three covers.
From here every post the campaign produces ships with its own custom cover, automatically. You save the equivalent of fifteen to twenty minutes per article that you used to spend hunting for stock photography — and your blog ends up with a more recognisable, more consistent visual identity than most human-curated sites manage.