DataForSEO Brain

AI Assistants as DataForSEO Surfaces

4 min read

AI Assistants as DataForSEO Surfaces

ChatGPT, Claude, Gemini, and Perplexity as answer surfaces that DataForSEO queries, scrapes, and monitors for brand mentions and citations. Sits under DataForSEO Brain then Platforms.

Overview

Generative engines are now a discovery surface alongside classic search, and DataForSEO exposes them through the AI Optimization API. The data layer has four parts: LLM Responses (send a prompt to a model and get a structured answer with citations), LLM Scraper (cheaper real-time collection of the consumer ChatGPT and Gemini interfaces), LLM Mentions (a queryable database of where brands and domains are cited inside AI answers), and AI Keyword Data (AI search volume derived from People Also Ask). This is raw infrastructure for Generative Engine Optimization (GEO), not a finished dashboard, so teams build their own tracking on top.

What it covers

  • LLM Responses for chat_gpt, claude, gemini, perplexity: structured answers with reasoning items, message items, source annotations[] (title, url), and fan_out_queries[]. ChatGPT, Claude, and Gemini support Live plus Standard; Perplexity is Live-only.
  • LLM Scraper for chat_gpt and gemini: scrapes the live consumer interface, returning markdown, search_results[], sources[], typed items (text, table, images, products), and brand_entities[].
  • LLM Mentions: five Live-only endpoints (search, top_pages, top_domains, aggregated_metrics, cross_aggregated_metrics) tracking mentions and citations in Google AI Overview and ChatGPT.
  • AI Keyword Data keywords_search_volume: ai_search_volume and ai_monthly_searches (12-month trend), computed from People Also Ask SERP data.

Key parameters / inputs

  • LLM Responses: user_prompt (required, up to 500 chars), model_name (required), max_output_tokens, temperature, top_p, system_message, message_chain (up to 10), and web_search / force_web_search / use_reasoning where the model supports them.
  • LLM Scraper: keyword (up to 2000 chars), one location field, one language field, force_web_search.
  • LLM Mentions: target array of up to 10 entities, each a domain (with search_scope sources/search_results) or keyword (with match_type word/partial and search_scope question/answer/brand_entities/fan_out_queries), plus platform (chat_gpt or google), location/language, initial_dataset_filters.
  • AI Keyword Data: keywords (up to 1000, max 250 chars each), one location field, one language field.

Response / what you get back

  • LLM Responses result: model_name, input_tokens, output_tokens, reasoning_tokens, web_search, money_spent (USD, the third-party model charge passed through), items[] (reasoning and message sections with optional source annotations), fan_out_queries[].
  • LLM Mentions search item: platform, model_name (for google always google_ai_overview), question, answer (markdown), sources[] (source_name, snippet, title, domain, url, position, publication_date), ai_search_volume, brand_entities[].
  • Aggregation endpoints group by location, language, platform, sources_domain[], search_results_domain[], brand_entities_title[], with mentions and ai_search_volume per group (impressions is deprecated and returns null).

Cost & method notes

  • Per the research report and live pricing pages: LLM Responses costs 0.0002 plus a 0.0012 per page Standard, 0.004 Live (approximate); LLM Mentions about 0.001 per row, roughly 0.01 per task plus $0.0001 per keyword.
  • LLM Responses, Scraper, and Mentions all run up to 120s with 30 concurrent Live tasks per platform; 2000 calls/min. See cap-rate-limits-throughput and cap-task-vs-live-execution.

When to use / how it fits

Gotchas / limits

  • Coverage is asymmetric: LLM Mentions covers Google AI Overview across multiple locations but ChatGPT only for United States (location_code 2840) and English. The launch set was Google AI Overview plus ChatGPT (US, GPT-5).
  • LLM Responses source annotations may return empty even when web_search is true; some models do not support web_search (for example o1, o3-mini).
  • Platforms overlap little in their citations, so each engine must be tracked separately.
  • Practitioner evidence (GEO paper, Ahrefs) shows quotations, statistics, cited sources, YouTube mentions, and branded web mentions correlate with AI visibility far more than backlinks; keyword stuffing does not transfer. See cap-geo-ai-search-optimization.
  • ChatGPT web answers lean on Bing’s index and Perplexity leans heavily on community discourse (Reddit), so the underlying source mix differs sharply per engine.
  • DataForSEO is raw infrastructure, not a SaaS dashboard, so prompt design, scoring logic, and share-of-voice math are the caller’s responsibility; finished trackers (Profound, Peec, Otterly) sit in a complementary category.

Sources

Graph View

Backlinks