Bring AnnuityRatesHQ into your AI conversation

Every numeric result includes a data_context object with its as-of time, universe, sample size when meaningful, provenance, canonical URL, and the JSON scope that context covers.

null means unavailable from the governed public source. It never means zero, false, or that a contract feature does not exist. If a source fails, the server returns unavailable; it does not present the failure as an empty successful answer.

Product detail includes a versioned contract_facts section for published issue and joint rules, accepted ownership and fund types, withdrawals, annuitization, fees, bonuses, surrender schedules, market-value adjustments, return-of-premium provisions, and source vintage. A section marked not_published is evidence of missing publication—not evidence that the feature is unavailable.

Product detail also includes scenario_facts for the requested state, premium, issue age, deferral period, and planned income-start age. Eligibility checks stay null when the public source cannot support a decision, and income-plus-legacy questions always retain an explicit carrier-illustration verification list.

For research_annuity_income_and_legacy, an optional query must exactly name one supported public FIA product, carrier, or canonical product URL. The public F&G brand spellings F&G, FG, and F G are treated as the same leading identity token; no broader partial matching is used. Focused mode resolves only through ARHQ's public FIA mappings and curation, uses authoritative product-pinned illustrations, obeys limit, and returns unavailable rather than substituting unrelated generic candidates. Product focus cannot initially be combined with comparison_income_start_ages.

Cite the most specific canonical_url returned with the fact: product pages for contract facts, carrier pages for carrier facts, and the rate-change page for observed history.

Pagination cursors are bound to the original filters. Changing the term, state, premium, rating, query, or sort while reusing a cursor returns an explicit error payload instead of mixing two result sets.

Use get_arhq_data_health when checking freshness or interpreting zero results. A healthy zero-event change feed is different from an unavailable source.

The current 18-series rate index is available through get_annuity_rate_index. Percentage rows include both a decimal-fraction value and percentage-point value_percent. Trend fields remain unavailable until same-cohort append-only history is trustworthy; a null delta or empty series never means no change.

For a specific MYGA term, call search_myga_rates. Use the broad search_annuity_products tool for discovery across product families, carrier names, riders, or index terms.

Call get_annuity_product before asserting contract-level availability, surrender terms, issue ages, premium ranges, riders, or state fields. Search results are discovery records, not full contract records.

For income-and-legacy questions, compare candidates under the same age, state, premium, and deferral period. Report the income mechanism, rider cost, liquidity, joint availability, death-benefit evidence, and source vintage separately; never treat a high payout as proof of stronger beneficiary value.

When the user names a product or carrier, pass its exact public identity as query to research_annuity_income_and_legacy. Contradictory qualifiers, unmatched URLs, ambiguous mappings, unavailable income curation, unproved rating floors, and ledgers without a complete requested-age election boundary followed by the first priced income period all remain unavailable.

When presenting a ladder, report its constraints and carrier_concentration, keep maturity dates separate, and surface alternatives when concentration or rating tradeoffs matter.

For market benchmarks, call get_annuity_rate_index, name the exact series, state its as-of time and cohort, and use value_percent for percentages. Never infer a trend from unavailable history.

A high-quality numeric answer states the scenario, source vintage, universe or sample size, factual result, canonical link, and the limitation most likely to change the interpretation.