To track which competitors AI search engines recommend, run your category questions and "best [category]" queries across ChatGPT, Gemini, Perplexity, Claude, and Grok, then log three things for every answer: which competitors appear, where they rank in the list, and crucially which sources the engine cites for each one. Add competitor-name queries to see how each rival is described, then benchmark all of it against your own presence to find the exact sources and pages you are missing. This guide walks through the method query by query, and how to run it on a repeating cadence so you catch shifts instead of snapshots.
This is competitor tracking, not brand tracking. If you only want to check whether your own name shows up, see how to check if your brand shows up in ChatGPT. Here the goal is to map the competitive field AI search engines present to your buyers, then reverse-engineer why your rivals are in it and you are not. That mapping is the start of real AEO work.
What does tracking competitors in AI search actually mean?
Tracking competitors in AI search means recording which rival brands AI search engines name when buyers ask category and "best of" questions, what position those rivals hold in the answer, and which web sources the engine cites to justify naming them. It is not about watching your own brand. It is about building a running picture of the competitive set AI presents and the citations propping it up.
The reason citations matter more than rankings is that AI search engines do not invent recommendations. They assemble them from third-party sources: Reddit threads, review sites, listicles, and editorial coverage. Across the engines our research tracks, the share of citations pointing to a brand's own domain runs from roughly 23% on ChatGPT down to about 6% on Claude, which means the overwhelming majority of what props up a competitor is coverage they do not own. If you can see the sources behind a competitor's placement, you can see exactly what you would need to match.
What to do: Treat every competitor answer as two data points, not one. Record the recommendation (who, where), and record the evidence (which URLs the engine cited). The second is the one that tells you what to build.
Step 1: Build your query list
Start with three groups of queries: category queries, "best [category]" queries, and competitor-name queries. Together they cover the three ways a buyer learns about your competitive field through an AI search engine, and each surfaces a different layer of competitor data.
Category and "best of" queries surface the recommendation set, the shortlist AI hands a buyer who is comparing options. Competitor-name queries surface the narrative, how the engine describes a specific rival and what it cites to back that description up. You want both, because a rival can dominate the shortlist while having a thin, attackable narrative.
Build the list like this:
- Category queries. The questions a buyer asks before they know any brand names. "What tools help agencies track AI search visibility?" or "Best way to monitor brand mentions in ChatGPT." Write these as full sentences with constraints, because that is how people query AI search engines, and longer queries surface a wider competitive set.
- "Best [category]" queries. The explicit shortlist requests. "Best AEO platform for ecommerce," "best AI visibility tool for small teams," "cheapest competitor monitoring for AI search." These map directly to the listicles AI search engines lean on.
- Competitor-name queries. One per rival. "Is [competitor] good for AI search tracking?" and "[competitor] alternatives." These reveal how the engine narrates each competitor and, in the alternatives query, who it groups them against.
Aim for breadth over volume at first. Ten to twenty well-constructed queries across the three groups will surface the bulk of your competitive set. Manual checks cap out fast, which is the constraint a monitoring platform exists to remove.
Step 2: Run every query across multiple engines
Run each query in a fresh, logged-out session on every AI search engine your buyers actually use, because the engines disagree on who to recommend far more than people expect. A competitor that owns position one on ChatGPT can be absent on Perplexity, and the citation behavior behind those answers differs just as much.
The engines do not behave alike. Our research found ChatGPT is the most volatile, changing roughly 39% of the brands it names between identical runs, while Claude is the most deterministic at around 8%. Reddit citations cluster heavily on Grok and Perplexity and barely register on Claude. If you only check one engine, you are not tracking your competitors in AI search. You are tracking them on one surface and guessing about the rest.
Use a clean session every time so your own history does not bias the answer. Run the same query across each engine back to back so you are comparing the same intent. Capture the full answer, not just whether your competitor appears, because position and the cited sources are the data you came for.
What to do: Pick the engines that match your buyers and your budget. Most teams start with ChatGPT, Gemini, and Perplexity, then add Claude and Grok if their category shows Reddit or developer-audience activity. Match the engine list to where your buyers actually ask.
Step 3: Log who appears, their position, and the sources cited
For every answer, record four columns: the engine, the competitors named, each competitor's position in the answer, and the exact URLs the engine cited for them. The sources column is the one most people skip, and it is the one that turns tracking into an action plan.
Position matters because AI search engines present recommendations as an ordered list, and the brand named first carries disproportionate weight. A competitor sitting at position one across four engines is in a different situation than one mentioned last on a single engine. Logging position over time also shows you movement, which is the whole point of tracking on a cadence.
The sources column is where the strategy lives. When you see that a competitor's position-one placement on Perplexity traces back to a specific Reddit thread, a G2 category page, and a "best of" listicle, you are no longer guessing why they win. You can read the exact pages doing the work. This is why AI citations come from third-party sites rather than brand pages: the engine is quoting the open web's verdict, and that verdict is now sitting in your spreadsheet as a list of URLs.
A simple log looks like this:
| Column | What to record |
|---|---|
| Engine | ChatGPT, Gemini, Perplexity, Claude, or Grok |
| Query | The exact query and which group it belongs to |
| Competitors named | Every rival brand in the answer |
| Position | Order each competitor appears in the recommendation |
| Sources cited | The exact URLs the engine cited for each competitor |
| Date | When you ran it, so you can compare across cycles |
Step 4: Benchmark against your own presence
Add your own brand to the same log, measured the same way, then compare row by row to find where competitors have citations and you have nothing. The gap is rarely product quality. It is source coverage: the review pages, threads, and roundups that mention them and not you.
Run your own brand through the identical query set and record your appearances, positions, and the sources cited for you, if any. Now you have an apples-to-apples comparison. If three competitors show up for "best AEO platform for ecommerce" and you do not, the next column tells you why: they are cited by a Capterra category page and two listicles your brand is missing from. The fix is not to rewrite your homepage. It is to earn presence on those specific sources.
How to fix this: Sort your log by the sources that cite competitors but never cite you. Those recurring domains are your priority list. A G2 category page, a particular subreddit, and the top "best of" article for your category will usually account for most of a competitor's lead. Earning accurate, current mentions on that short list of sources is the highest-impact work, far more so than publishing another page on your own site. For the deeper reasoning on which signals drive placement, see how ChatGPT decides what to recommend.
Step 5: Repeat on a cadence
Re-run the full query set at least monthly, because AI search results change and a single snapshot measures noise, not your standing. The same query can name a different competitor set on different days, and engines refresh which sources they trust as new content enters their 30-day retrieval window.
A snapshot tells you who showed up once. A cadence tells you the trend: whether a competitor is gaining position, whether a new rival just entered the set, whether the Reddit thread that anchored someone's lead has dropped off, and whether your own presence is climbing as your source coverage grows. Recency is a primary retrieval signal for AI search engines, so the competitive field genuinely shifts month to month, and only repeated measurement catches it.
Run it monthly at minimum. Weekly is better if your category moves fast or you are actively building presence and want to see it land. Keep every cycle in the same log so you can chart position and source changes over time. For the full set of metrics worth charting, see how to measure AI search visibility.
Doing this continuously at scale
Manual competitor tracking works for a handful of queries across two or three engines, but it breaks down at the scale where the data gets useful: dozens of queries, five engines, every competitor, repeated monthly, with sources logged for each. That is hundreds of answers to capture and classify per cycle, by hand, before you have done any analysis.
This is the work a monitoring platform automates. The Loudmink AEO platform runs your query set across up to five AI search engines on 24-hour cycles, logs which competitors appear and where they rank, and classifies the sources behind every answer so you can see exactly which third-party pages drive your competitors' placements and where your own coverage is thin. It tracks 50 to 300 queries depending on tier, which is the volume manual checks cannot sustain. Plans from $99/mo as of June 2026. You can see your starting position with a free scan before committing to anything.
Loudmink turns this into ongoing competitor and visibility tracking across AI search engines, with the source intelligence that shows where each recommendation comes from. Plans from $99/mo.
Frequently Asked Questions
How do I find out which competitors ChatGPT recommends?
Open a fresh, logged-out ChatGPT session and ask the category and "best [category]" questions your buyers ask, not your own brand name. Record every competitor named, the order they appear in, and the sources ChatGPT cites for each. Repeat across the other AI search engines, because ChatGPT's recommendations often differ from Perplexity's or Gemini's.
Why should I track the sources AI cites, not just the rankings?
Because the sources are the cause and the ranking is the effect. AI search engines build recommendations from third-party pages like Reddit threads, review sites, and listicles, with most citations pointing to sources a brand does not own. Logging those URLs shows you the exact pages propping up each competitor, which is the list of places you need presence to compete.
How often should I track competitors in AI search?
Monthly at a minimum, weekly if your category moves quickly. AI search results change from day to day and engines refresh which sources they trust as new content enters their retrieval window, so a single snapshot measures noise. Only repeated tracking in the same log reveals whether a competitor is gaining position or a new rival has entered the set.
Can I track competitors across AI search engines manually?
Yes, for a small set. A spreadsheet, ten to twenty queries, and two or three engines is enough to map your competitive field once. It stops scaling when you need every competitor, all five engines, sources logged for each answer, and the whole set re-run monthly, which is where an AI search monitoring platform takes over.
Do different AI search engines recommend different competitors?
Yes, often. The engines disagree on who to recommend far more than buyers expect, and their citation behavior differs too, with Reddit citations clustering on Grok and Perplexity while ChatGPT links to brand sites more than the others. Tracking only one engine gives you a partial and sometimes misleading view of your competitive set.