Best Practices

Patterns that consistently produce strong AI influencers and high-performing content on AutoPersonas.

Writing effective influencer personalities

Start with a clear archetype, fitness coach, tech reviewer, lifestyle minimalist, etc.
Big Five traits: pick 2-3 dominant traits and leave others neutral.
Interests should be specific, not just “food”, but “Italian home cooking”.
Quirks make influencers memorable, add 2-3 distinctive behaviors.
Catchphrases, 1-2 maximum, used sparingly.

Import first. Run Import Posts before you generate anything, imported captions seed the voice corpus that grounds every future draft in a real published voice.
Study the account's actual captions and match their register. If the real voice is terse and blunt, generated captions should be too, resist the urge to “polish” it into generic influencer-speak.
Keep captions evergreen. Avoid date, holiday, and travel hooks that expire; they age badly when a post sits in the queue or gets repurposed later.
Each caption should be distinct, not a reworded copy of a previous one. The duplicate detector blocks exact repeats and warns on near-duplicates, but variety is still on you.

Every generated post lands in Pending Review. Approve, edit, regenerate, or reject, then schedule. Publishing only happens after approval.
Review everything manually for a new influencer's first 10-20 posts. Only consider auto-approve once the output is consistently on-voice.
Prefer Regenerate over hand-rewriting a bad caption, and use Repurpose to re-run proven posts instead of generating from scratch.
When generating, include a specific mood, a setting, and a time of day, vague briefs produce generic posts.

The influencer editor's Generation tab controls how posts are styled on top of the identity from the wizard:

Feed style presets, Balanced, Creator-forward, Lifestyle / editorial, and Unpolished / real bundle curated per-archetype weights. Start from the one closest to your niche.
Voice style presets, Balanced, Terse / signature, Warm & conversational, and Expert / product-focused do the same for caption archetypes.
Open Adjust feed-style mix / Adjust voice-style mix to tweak two or three per-archetype weights directly, higher weight means the scheduler picks that archetype more often, and 0 drops it entirely. Then save a custom preset so you have a reusable baseline without hand-tuning every slider again. Use Suggest more to draft style keywords and catchphrases from the influencer's persona.
Leave the QA pass on for new influencers, it scores every generated image against a realism and identity rubric (identity drift, hand anatomy, garbled background text, stray brand logos, archetype fit) and automatically regenerates failures. An image that still fails after retries is not discarded, it lands in your review queue flagged for human review, so you make the final call.
Pair narrative mode serialized with life moments (Lifestyle step) for the strongest post-to-post continuity; use episodic for editorial, self-contained feeds.

The built-in pipeline plus presets covers most rosters, see Custom Flows (Beta, Pro and Agency plans) for when a flow is warranted and what nodes exist.
Use a free dry run to verify wiring before a billed test run.

Engagement engines (comment responder, growth engine, DM manager) are available on Pro and Agency plans. Drafted replies wait in the approval queue until you approve them.
Keep manual approval on for the first few weeks; enable per-engine auto-approve only after the drafts consistently match your voice.
Built-in safety filters skip sensitive conversations (politics, religion, tragedy, minors, advice topics) automatically, don't try to route around them.
On X, reply to conversations in your niche early; thoughtful replies out-perform broadcast posting for new accounts.

Generate multiple caption variants for the same image.
Publish variants across different posting times.
Use the Learning Dashboard on the Analytics page to see what works, then iterate on high-performing patterns.
Repurpose your best Published posts onto a fresh schedule instead of always generating new material.