Vendor Tools · Vendor Tool

LLMrefs

A dated analyst profile of LLMrefs, a keyword-first AI visibility tracker with one All-in-One plan and automatic prompt fan-out.

LLMrefs is a commercial AI search analytics vendor. On 2026-09-07, its public site sold a keyword-first workflow: import keywords, generate fan-out prompts, measure brand mentions, citations, share of voice, and position across a long list of AI search engines, then export CSV or pull the same objects through an API.

Geolyze files it with AI visibility tools because the measurement objects are keywords expanded into prompts, generated answers, citations, competitors, and share of voice. It is not a classic SEO rank tracker.

This profile is an analyst file from official marketing pages and the published OpenAPI spec captured on 2026-09-07, with the pricing block re-captured on 2026-10-26. It is not a hands-on product test. Live dashboard behavior stays unknown.

Best fit

Use LLMrefs when the team already thinks in keywords, does not want to hand-author every prompt, and can accept a single All-in-One bill that names many engines with no per-engine add-on fee.

Unlimited seats and projects fit an agency that will report several clients from one subscription. Weekly dashboard updates are a better match when the operating rhythm is a weekly review, not an intra-day alert loop.

It is a weaker fit when extra-keyword pricing must be public before purchase, when SSO or SOC 2 must be named in packaging, or when official cadence copy has to be unambiguous.

What official pages claim

Public pages name these modules. Treat them as vendor-stated capabilities, not Geolyze lab results.

The OpenAPI keyword payload includes prompt text and answer text. That is the strongest public evidence that LLMrefs stores the objects a GEO checklist needs. Geolyze has not called the API. If a demo only shows a score, ask for the fan-out prompts and the dated answer.

Platform coverage

Named engines in this capture:

• ChatGPT (FAQ copy said ChatGPT GPT-5)
• ChatGPT Search
• Google AI Overviews
• Google AI Mode
• Gemini
• Perplexity
• Claude
• Grok
• Microsoft Copilot
• Meta AI
• DeepSeek

The public All-in-One plan said all AI search engines with no additional fees. Whether every account and every country actually includes every named engine is unknown.

Geo targeting was 20+ countries and 10+ languages on one product page, and 50+ countries and 20+ languages on the pricing block. Keep that conflict; do not pick a number.

Refresh copy also conflicts. One product page said prompts run across engines daily. The homepage FAQ said checks happen in real time and that every keyword updates in the dashboard at least weekly. Cadence beyond “at least weekly on the dashboard” stays needs_review.

Pricing note

Captured 2026-09-07 and re-captured 2026-10-26 from https://llmrefs.com; the /pricing URL renders the same page with an embedded pricing section.

7-day free trial. No credit card required to create an account. Cancel anytime. Annual billing was not shown on either capture.

The same FAQ still asked about needing “more than 50 keywords.” That line sits beside the 500-prompt plan cap. Treat keyword versus prompt units as unresolved. Default authenticated API rate limit in the OpenAPI spec was 10 requests per minute.

Re-check live packaging before a purchase decision. A price labeled limited time is not a stable SKU; it has now carried that label across two captures seven weeks apart.

Strengths

One public price with a long named engine list and unlimited seats is easy to compare. Many category peers gate engines by plan or sell them as add-ons.

The keyword-first model matches how SEO teams already collect demand. Fan-out can be an advantage if the team lacks a prompt taxonomy; it is a quota risk if each keyword expands into many prompts.

The published API schema includes share of voice, sources, prompt text, and answer text. That is more than a marketing score.

Limitations

Official pages disagree on geo breadth, refresh cadence, and whether the commercial unit is a keyword or a prompt. Extra volume pricing is unpublished. Getting-started docs were a placeholder.

SSO and SOC 2 were not found on the fetched official pages. Vendor case-study share-of-voice percentages, “10,000 marketers,” and unsourced zero-click statistics are not Geolyze facts. Live weighting methodology and dashboard quality remain unknown.

Example and counterexample

A workable use is an in-house SEO importing a few dozen category keywords, reviewing weekly share of voice and cited domains, and exporting CSV for a monthly report.

A mismatch is treating “500 prompts” as 500 keywords, then hitting a lower keyword cap implied by the “50 keywords” FAQ, or assuming every named engine is proven on day one of the trial.

Alternatives

The published Geolyze vendor profiles at the time of this page are AIvsRank, Profound, Peec AI, Scrunch, Goodie, Otterly.AI, and AthenaHQ. Use AIvsRank when the next job is measurement, benchmarking, or audit rather than LLMrefs’ keyword-first tracker; the others are compared with dated captures in Best AI Visibility Tools.

Other AI visibility vendors should not be compared feature-by-feature here until they have their own reviewed profiles.

For buying criteria without a vendor shortlist, use Best AI Visibility Tools and How to Measure AI Visibility. Related terms: prompt tracking, AI visibility, AI share of voice.

Operator checklist

1. Decide whether a keyword-plus-fan-out model is enough, or whether the team must author every prompt.
2. Ask how a keyword consumes the 500-prompt quota.
3. Ask whether every named engine is on from day one, including trial.
4. Ask for a dated sample of keyword, fan-out prompts, full answer text, and citations. A score without answer evidence is not enough.
5. Confirm geo markets against the live UI, not the conflicting marketing counts.
6. Re-check the “limited time” $79 price. This note is from 2026-10-26.

Common misunderstanding

LLMrefs says you track keywords, not prompts. In Geolyze terms, the product is still prompt tracking underneath: fan-out turns each keyword into several prompts, and the public plan cap is 500 prompts. Buying “keyword tracking” does not mean prompts are unlimited.

Verification note

Official pages were captured on 2026-09-07 and the pricing block was re-captured on 2026-10-26 in a live browser. The price, quota, engine claim, and trial terms were unchanged; the geo-count conflict inside the page persists and is recorded with both readings. Geolyze has not operated an LLMrefs workspace. Recapture the All-in-One price, prompt quota, engine list, and keyword-unit language before relying on this page for procurement.