Can I track competitors in AI search for free?

Yes, by hand, and Prefer's 7-day trial is free with no card. By hand, pick 20 to 30 buyer questions, ask each engine, and record which brands get named and which sources get cited in a spreadsheet. It is tedious to repeat weekly across five engines, which is why teams move to a tool once the habit sticks.

Why track competitors on more than one engine?

Because the engines barely agree. In Prefer's study only 5 of 713 cited domains appeared on all four engines, and about three-quarters were cited by just one. A competitor who dominates Perplexity may be absent from ChatGPT, so single-engine tracking gives a false read.

What should I measure, mentions or citations?

Both, per competitor. Prefer logs both for every competitor you track: how often each rival is named in the answer (share of voice) and which of their pages get cited as sources. Mentions tell you who AI recommends; citations tell you which content earned it, which is where you find your gap.

How do I track competitors in AI search?

Run the same buyer prompts across every AI engine and log who gets named and cited. Prefer does it across five engines, with share of voice.

Javed Khatri Co-founder, Prefer

5 September 2026 Updated 2 October 2026 2 min read Answer

The short answer

How can I see which competitors AI recommends?

Run the same buyer questions across the AI engines on a schedule and log who each answer names and cites. Prefer does this on ChatGPT, Gemini, Perplexity, Google AI Overviews and AI Mode, with share of voice per competitor. Compare every engine: their sources barely overlap.

Key takeaways

In Prefer's four-engine study, only 5 of 713 cited domains were shared by all four engines and about 75% were cited by a single engine; ChatGPT drew on 89 distinct domains, Claude 220, Perplexity 273, and Gemini 367, so tracking one engine measures roughly a quarter of the picture. (Prefer AI search citation statistics)
Prefer's 2026-09-02 ChatGPT run found a recurring competitor shortlist (Profound, Peec AI, Semrush, Ahrefs, Otterly, Scrunch), with Profound named first in about 12 of 16 category prompts, measured by running buyer prompts and reading each answer's named brands and sources. (Prefer AEO loop gap report, run 1)
AI visibility tools track competitors as a metered feature: Prefer's plans include 3, 10, and 30 tracked competitors by tier, with share-of-voice comparison per prompt. (Prefer pricing)

The method: same prompts, every engine, on a schedule#

Track competitors by running a fixed set of buyer questions across every engine on a schedule and logging who each answer names and cites; Prefer does this across five engines. The method is the same whether you do it by hand or with a tool:

Build a prompt set. List 20 to 30 questions your buyers actually ask, category questions (“best [category] tool”), comparison questions (“[you] vs [rival]”), and problem questions. These are where recommendations get made.
Ask every engine. Send each prompt to ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews. Running one engine is not enough, because the engines barely agree on sources.
Log two things per answer. Record which brands get named (share of voice) and which domains get cited as sources. Do this per competitor, per prompt.
Repeat on a cadence. Answers drift, so a weekly or monthly re-run turns a snapshot into a trend.

Prefer’s own ChatGPT run shows what this surfaces: a recurring competitor shortlist (Profound, Peec AI, Semrush, Ahrefs, Otterly, Scrunch) emerged from reading the named brands across prompts, with Profound named first in about 12 of 16 category questions. You only see a leaderboard like that by reading the actual answers, not by guessing.

Why one engine misleads, and where tools help#

The engines cite barely-overlapping sources, so single-engine tracking gives a false read. In Prefer’s four-engine study, only 5 of 713 cited domains were shared by all four engines, and about 75% were cited by a single engine and no other. The pools are different sizes too: ChatGPT drew on 89 distinct domains, Claude 220, Perplexity 273, and Gemini 367. A rival who owns Perplexity can be invisible on ChatGPT. Track one engine and you measure roughly a quarter of reality.

You can do all of this in a spreadsheet, and for a first pass you should. It gets painful fast: re-running 30 prompts across five engines every week, normalizing domains, and diffing week over week is a lot of manual work. That is why competitor tracking is a metered feature in AI visibility tools. Prefer’s plans include 3, 10 and 30 tracked competitors by tier and compute share of voice per prompt automatically, so the leaderboard and the source gaps update on their own.

Whichever route you take, measure both halves: who AI recommends (mentions) and which of their pages earned it (citations). The citations are where you find the content gap to close. Our benchmark competitors workflow and competitive AI analysis module do this tracking for you across every engine. To see your competitive share of voice, run a free AI visibility audit.

Sources