# Reply (Search) Searches X for tweets matching your queries and replies with templates or AI-generated text. The primary growth engine for small accounts. ## How it works Each cycle, the module shuffles your configured search queries (randomized order to avoid detection), searches X for matching tweets, filters by age and engagement, and posts replies. Queries are picked randomly each cycle. If a query has a per-query cap, replies stop for that query once the cap is hit, then move to the next. ## Search queries Queries are configured as an object mapping search strings to reply content. Each query can use either a static template or AI generation: - **Static template:** `"search term": "Reply text here"` - **AI generation:** `"search term": { "use_ai": true, "prompt": "Your AI prompt" }` Individual queries can be temporarily disabled via the Skipped Queries list. If all queries are skipped, the module waits for the next cycle. ## Configuration reference | Setting | Default | Description | | **Search Queries** | — | Query-to-reply mapping (see above) | | **Skipped Queries** | none | Queries to temporarily disable | | **Pages Parsing** | 10 | X Search pagination depth per query (max 30) | | **Post age (min)** | 60 | Only reply to tweets younger than this many minutes | | **Replies per query** | 0 (no limit) | Max replies per query per cycle | | **Delay between replies (sec)** | 5 | Base delay between replies (+ random 0-10s jitter) | | **Replies per day** | Plan-driven | Daily cap (overridden by plan) | | **Delay between cycles (min)** | 1 | Wait between full cycles (min 0.5 min) | | **Author followers** (min) | 0 | Skip tweets from authors below this follower count | | **Author followers** (max) | none | Skip tweets from authors above this count | An additional random pause of 30-120 seconds is injected roughly every 30% of replies as an anti-pattern measure. ## Daily caps | Plan | Daily limit | Scope | | Free | 30 | Reply (Search) only | | Standard | 100 | Reply (Search) only | | Pro | 300 | Combined with Reply (List) | | Unlim | Unlimited | Combined with Reply (List) | Caps reset at UTC midnight. On Pro and Unlim, Reply (Search) and Reply (List) share one combined counter. ## AI text generation When a query is configured with AI enabled, the module sends the target tweet's text to the AI provider (default: Grok) along with your custom prompt. The AI response is validated before posting: - Empty responses are rejected. - Responses that are only a URL are rejected. - Max text length is 700 characters by default (configurable in AI Settings). ## Advanced filters (Unlim only) These settings are only available on the Unlim plan. On lower plans they are locked and use safe defaults: | Setting | Default | Description | | **Wallchain X Score** | 0 | Minimum X-Score (0-1000) to reply. Scores are cached for 1 hour. | | **GEO blacklist** | none | Comma-separated country codes to exclude. | | **Blue verified** | off | Only reply to verified accounts. | | **Follow author** | off | Follow the tweet author after replying. | | **Follow author probability** | 100% | Probability to follow when enabled. | | **Skip my followers** | off | Skip authors who already follow this account. | | **Like author post** | off | Like the tweet after replying. | ## Safety behavior - Never replies to the same tweet twice. - Skips self-authored posts. - GEO filtering is fail-closed: if the country lookup fails, the tweet is excluded. - Wallchain X-Score validation is also fail-closed on API error. - Anti-spam context is passed with each reply to make replies appear organic to X. Related reading: [Reply automation safety](https://helperx.app/blog/reply-automation-safety) · [AI replies that don’t sound like AI](https://helperx.app/blog/ai-replies-that-dont-sound-ai) · [GEO blacklist setup & ethics](https://helperx.app/blog/geo-blacklist-reply-search-setup) · [Wallchain X-score filters](https://helperx.app/blog/wallchain-x-score-filters-explained).