How to track LLM citations (with a 2026 tools comparison)

Tracking LLM citations is now part of standard SEO measurement in 2026. You measure your AI-search visibility the same way you measure Google rankings. The mechanics are different (you don't have a SERP tracker; you have a prompt audit), but the goal is the same: know whether your content is being surfaced when your buyers ask AI engines the questions you'd want to be the answer to. This piece walks through what to measure, the tools that do it well in 2026, and how to build a manual baseline if you can't spend on a vendor yet.

#What citation tracking actually measures

When an AI engine (ChatGPT, Claude, Perplexity, Google AI Overviews) generates an answer, it does one of three things with respect to your domain:

1. Cites you. Links to a specific URL on your domain in the answer's source list
2. Mentions you. References your brand by name in the answer text without linking
3. Absent. Neither cites nor mentions you

The hierarchy: cited > mentioned > absent. Citation is the goal because it's the form most likely to drive click-through. Mention is a partial win. It builds brand recognition but doesn't directly drive traffic. Absent is the diagnostic case: you're not in the model's preferred sources for that prompt.

The tracking question: across the prompts your ICP would ask, what's the distribution of cited / mentioned / absent? And how is that distribution changing over time?

#What to track

Three things, in priority order:

1. A defined set of prompts. 8–25 prompts that map to your highest-intent buyer queries. These should be conversational, specific, and the kinds of questions a real founder would type into ChatGPT or Perplexity. “Best productized agency for a B2B SaaS MVP” is a tracked prompt; “saas mvp” isn't (it's a Google query, not an AI engine query).

2. A defined set of AI engines. At minimum ChatGPT, Claude, Perplexity, Google AI Overviews. Optionally add Gemini standalone, You.com, Phind, depending on your audience.

3. Citation status per (prompt, engine) pair. Cited, mentioned, or absent. Optionally: rank position within the citation list, which page of yours was cited, what the answer text said about you.

The minimum viable tracking sheet: 10 prompts × 4 engines = 40 cells, populated monthly. Cells contain a tri-state (cited / mentioned / absent) and an optional notes column. Total time: ~1 hour per month if done manually.

#The tools

A non-exhaustive scan of the citation-tracking vendor landscape in 2026:

#Profound

The most comprehensive enterprise option in 2026. Tracks across 8–10 AI engines, supports prompt sets in the hundreds, integrates with GA4 for traffic correlation. Pricing starts around $499/month with custom enterprise tiers above that.

Best for: agencies tracking citations for 5+ clients, or large B2B SaaS with mature SEO operations.

#Athena (formerly Athena Insights)

Athena is the citation-tracking tool most commonly recommended in B2B SaaS circles in 2026. Cleaner UX than Profound, comparable feature set on the core tracking. $299/month for the starter tier, $599 for pro.

Best for: in-house marketing teams at funded SaaS companies who want a managed solution.

#Otterly

The most accessible option for early-stage teams. $49–$199/month tiers. Tracks the core 4 AI engines with a defined prompt set. Less polished UI than Profound or Athena but the price-to-signal ratio is good.

Best for: bootstrapped or pre-seed B2B SaaS, agencies experimenting with the category, founders who want a programmatic monthly audit without the enterprise sticker.

#Peec

Newer entrant with a free tier (limited prompt count) and paid tiers from $99/month. Strong on Perplexity tracking specifically; weaker on Claude and Google AI Overviews coverage.

Best for: teams whose ICP heavily uses Perplexity (developer tools, technical B2B), or teams piloting citation tracking before committing to a paid tier elsewhere.

#Honorable mentions

• HubSpot AEO (rolled out late 2025, $200/month add-on to existing HubSpot subscriptions). Tied to your CRM, useful if you're already in the HubSpot ecosystem
• Semrush AI Toolkit. Added to the existing Semrush platform, included in Pro plans
• Ahrefs AI Search reports. Rolled out in 2025, included in higher Ahrefs tiers

#The manual baseline (free, hour per month)

If you can't spend on a vendor yet, the manual approach is fine and arguably better as a starting point because it forces you to actually look at the answers AI engines are giving.

The setup:

1. Pick 8–10 prompts. These are the conversational queries your ICP would type into an AI search engine. Examples for a B2B SaaS dev agency: “best productized agency for SaaS MVP”, “fixed-price SaaS development agency under $20k”, “how much does a SaaS MVP cost in 2026”, “productized agency vs freelance developer for SaaS”.
2. Pick 4 engines. ChatGPT (free or paid tier), Claude, Perplexity, Google AI Overviews (just google your prompt; the AI Overview is the box at the top).
3. Run each prompt on each engine, monthly. Log the result in a spreadsheet: cited (with which URL), mentioned (with what text), absent.
4. Track the trend. Month-over-month, is your citation share growing? Which prompts are unchanged? Which are getting worse?

The whole exercise takes 60–90 minutes if you don't fall into a rabbit hole reading the answers. Do it on the same day each month for consistency. Save the answers as screenshots. AI engines are non-deterministic, so the same prompt next month may give a different answer, and you'll want the receipts.

A simple Google Sheets template:

Prompt | ChatGPT | Claude | Perplexity | Google AI | Notes
P1     | Cited   | Absent | Mentioned  | Cited     | ...
P2     | Absent  | Absent | Absent     | Mentioned | ...
...

Update monthly. Compute aggregate cited% per engine and per prompt. Watch the trend.

#What to do with the data

Citation tracking only matters if you act on it. The patterns to watch:

Engine-level patterns. If you're cited in Claude but not Perplexity, the diagnostic is usually that Claude has been trained on data sources you're well-represented in (likely your own site or sites that link to you), while Perplexity weights more recent or different sources. Different engines have different citation logic; you may need different content strategies per engine.

Prompt-level patterns. Prompts where you're absent across all engines are the priority. These are usually fixable with targeted content. Write a pillar post that's the literal best answer to the prompt, structure the answer in the first 60 words, add FAQ schema, and watch the citation status change over 4–8 weeks.

Prompt-level wins. Prompts where you're cited across 3+ engines are the wins to consolidate. Make sure those pages are well-internally-linked, schema-validated, refreshed quarterly. Defending an existing citation is cheaper than earning a new one.

Competitor patterns. Track competitors on the same prompt set. When a competitor enters citations on a prompt where you were absent, look at what they shipped. Usually a new pillar post or a comparison page. Mirror or counter.

#What changes citation rates

A non-exhaustive list of things that shift citation rates over 4–12 weeks of effort:

• Answer blocks at the top of pages (a 40–60 word literal-prose answer in the first 60 words of every money page). Single biggest unlock per hour of work.
• FAQPage schema with conversational Q&A pairs. AI engines extract these directly.
• Pillar-and-spoke topical clustering. A pillar page with 5–15 spoke posts linking back to it builds topical authority that engines can attribute.
• Comparison content. “X vs Y for Z” posts get cited disproportionately because they map to a comparative-question shape that AI engines surface.
• Reddit and HN mentions. Both are heavily-trained-on sources. A few authentic Reddit threads where your brand is mentioned in context can move citation rates.
• Backlinks from citable sources. Industry-standard authoritative sites (DEV.to, Smashing Magazine, LogRocket) being linked from yours and linking back lifts citation rates over months.
• Recency. AI engines weight recent content. Pages refreshed quarterly maintain citations 3x longer than pages that go stale.

For the comprehensive playbook on these levers, see AEO vs SEO: what changed in 2026 and the minimum schema markup every SaaS homepage needs in 2026.

#What we ship for clients

For our AEO Retainer engagements in 2026, the default citation-tracking stack:

• Tool: Otterly for early-stage clients, Athena for funded SaaS, Profound for enterprise
• Prompt set: 15–25 prompts mapped to the client's high-intent buyer queries
• Cadence: monthly automated reports + a manual sanity-check on the top 5 prompts
• Reporting: month-over-month citation share per engine, per prompt; competitor benchmarks; trend graphs
• Action triggers: any prompt where citation status drops month-over-month gets a content review; any new prompt where a competitor enters citation gets a counter-content task

That setup runs ~$300–$600/month in tooling for the client; the engagement value is in the action triggers. The “what we should do about this” layer that the tools themselves don't provide.

#What changes the tooling landscape in 2026

Watch for:

• Native AI engine analytics. Both OpenAI and Anthropic have hinted at first-party citation analytics for content owners. If they ship this, third-party tools partially commodity. Likely 2026–2027 timeline.
• Schema.org additions for AI. Google's been iterating on schema types specific to AI Overviews. New types could change what's worth marking up.
• Tool consolidation. The category will consolidate to 3–5 winners. Pick a tool that fits your scale; expect to revisit the choice in 12–18 months.

#Bottom line

Citation tracking is a fundamental marketing measurement in 2026, not a niche thing. The manual baseline is fine to start; the tools are worth their cost once you're investing in AEO content actively. The action layer matters more than the tracking layer. Track to know what to fix, then fix it.