DataForSEO Brain

LLM Mentions and Visibility

4 min read

LLM Mentions and Visibility

Structured data on where brands, domains, and keywords are mentioned and cited inside AI answers. Sits under DataForSEO Brain → Concepts.

Overview

The LLM Mentions API (launched 2025-11-11) is DataForSEO’s flagship AI-visibility tracker. It answers the question traditional SEO tools cannot: is my brand named, and is my site cited, inside generative AI answers. It maintains a mentions database collected from Google AI Overview and ChatGPT, each mention being a question-answer pair plus the sources the model cited. It is the measurement half of GEO, paired with the live-querying endpoints in cap-ai-optimization-api. All five endpoints are Live-only under /v3/ai_optimization/llm_mentions/.

What it covers

  • Search (/llm_mentions/search/live): raw mention records for target keywords/domains, returning question, answer (markdown), sources[], search_results[], ai_search_volume, monthly_searches[], brand_entities[], and fan_out_queries[].
  • Top Pages (/llm_mentions/top_pages/live): mentions metrics grouped by the most-cited pages for a target.
  • Top Domains (/llm_mentions/top_domains/live): mentions metrics grouped by the most-cited domains.
  • Aggregated Metrics (/llm_mentions/aggregated_metrics/live): consolidated metrics for one target across location, language, platform, and source domain.
  • Cross Aggregated Metrics (/llm_mentions/cross_aggregated_metrics/live): the same metrics for 2-10 targets at once, labelled by aggregation_key, for side-by-side brand or keyword comparison. Note the task spec slug cross_model_comparison maps to this real path.
  • Aggregation endpoints accept initial_dataset_filters with operators =, <>, in, not_in, like, not_like, ilike, not_ilike, match, not_match, applied before grouping.
  • links_scope (sources or search_results) on Top Pages and Top Domains controls whether grouping uses cited sources or retrieved search results.

Key parameters / inputs

fieldnotes
targetarray of up to 10 entities, each a domain or a keyword
domainentity; ⇐ 63 chars; no https:// or www; include_subdomains optional
keywordentity; ⇐ 250 chars; match_type word_match or partial_match
search_filterinclude or exclude (default include)
search_scopedomain: any/sources/search_results; keyword: any/question/answer/brand_entities/fan_out_queries
platformchat_gpt or google (default google; model then always google_ai_overview)
location_* / language_*default 2840 (US) / en; ChatGPT is US/English only
initial_dataset_filtersup to 8 filter expressions applied before aggregation
targets / aggregation_keyCross Aggregated only; 2-10 targets, key labels each group

Response / what you get back

Search returns total_count, current_offset, search_after_token, and items[]. Each item carries platform, model_name, question, answer, sources[] (source_name, snippet, title, domain, url, position, publication_date), search_results[], ai_search_volume, first_response_at, last_response_at, and brand_entities[]. The aggregation endpoints return a total object plus grouped arrays: location[], language[], platform[], sources_domain[], search_results_domain[], brand_entities_title[], brand_entities_category[]. Group elements carry key, mentions, and ai_search_volume (note impressions is deprecated and returns null).

Cost & method notes

  • Per the AI Optimization page, LLM Mentions is pay-as-you-go: roughly 0.001 per row, about $1.10 per 1,000 rows, with no monthly minimum since DataForSEO’s 2026-07-01 pricing update.
  • All endpoints are Live-only with one task per call. See cap-queue-priority-cost-model and cap-task-vs-live-execution.

When to use / how it fits

LLM Mentions is the data engine of play-ai-visibility-tracking and the GEO loop in cap-geo-ai-search-optimization. Use Search for evidence (the actual question and answer), Top Domains/Pages to see who AI cites in your category, and Cross Aggregated Metrics to compute Share of Voice across competitors. Practitioner research notes that about 85% of brand mentions come from third-party pages, so the off-domain sources_domain[] view matters as much as your own. The answer surfaces are profiled in plat-ai-assistants and the models in ent-llm-model-providers.

Gotchas / limits

  • Access is now plain pay-as-you-go, with no subscription or activation step. It returned 40204 through 2026-06-26; that gate was removed 2026-07-01 when DataForSEO moved all APIs to pay-as-you-go. A live re-probe on 2026-07-08 returned 20000 Ok with real data for /v3/ai_optimization/llm_mentions/search/live; fixture: .raw/sources/dataforseo-research/ai-optimization/fixtures/llm-mentions-search.json. See cap-status-error-codes.
  • ChatGPT coverage uses model GPT-5 and covers United States (location_code 2840) and English only; Google AI Overview (model_name always google_ai_overview) supports multiple locations and languages and is the broad-coverage platform. The launch announcement notes initial coverage is narrower than the marketing list.
  • Execution up to 120 seconds; 2000 calls/min; one task per call; offset max 9,000 (use search_after_token beyond 20,000 results).
  • Citation churn is high (industry studies report 40-60% of cited sources change month to month), so trends matter more than any single snapshot. See cap-data-collection-methodology.
  • For the google platform the model_name is always google_ai_overview; there is no per-Gemini-version selection here, unlike LLM Responses.
  • impressions is deprecated and returns null across all aggregation endpoints, so do not build metrics on it.

Sources

Graph View

Backlinks