Most teams start their competitor list the same way, by copying the companies their sales team runs into every week. That list is a fine place to begin, but it often doesn't match the brands AI tools actually name when a buyer asks for options. This guide shows you how to build a short list for competitors AI search tracking from what shows up in real answers, keep publishers on a separate list, and stop at a number your team can act on. By the end you'll have a shortlist of brands, a watchlist of source sites, and a record of where each name came from.
You'll need a rough picture of your buyers, a spreadsheet, and access to the AI tools your buyers are likely to use.
1. Define the market and the buyer decision
Start by writing down what you sell, who buys it, and which market you serve. Keep it to the unit a buyer would recognize, such as one product category or one use case, and not the whole company. Add the regions or languages that matter, because those change what AI tools say.
You're done when a teammate can read your notes and tell which buyer questions belong in the sample and which don't. A few sentences is enough.
The usual mistake here is treating every company in your CRM, or every SEO rival, as an AI search competitor. A company that sells to a different buyer can show up in a loosely related answer without being someone you compete with for the sale.
2. Draft a set of buyer prompts
Build a starter set of about 20 to 25 prompts. One Semrush guide to which AI search prompts to track suggests starting at around 25, and that number is about prompts, not competitors. Mix a few kinds of questions:
- Category questions with no brand in them, like "what are the best tools for [use case]?"
- Use case or constraint questions, like "what should a [buyer type] look for in a [category]?"
- Comparison questions, like "[Brand A] vs [Brand B]" or "alternatives to [Brand]."
- Decision stage questions that reflect the real options and criteria your buyers weigh.
Let your buyers' own words guide the phrasing. If three versions of a question mean the same thing, pick the one or two buyers are most likely to type, so your set isn't a pile of near-duplicates. If you're not sure where to start, our guide to the four types of prompts to track is a good map, and the keyword research for AEO guide shows how to turn a keyword list into questions.
DeepSmith can help with this step. Discover Prompts suggests starter prompts from your product, persona, and buyer-stage context in Deep IQ, and you choose which ones to add. It suggests prompts, so you still decide the competitor list yourself.
You're done when the set covers your main buyer situations and your comparison language. Label each prompt with its category and buyer stage so you can look at them in groups later.
Be careful not to write only prompts that name your own brand or the competitors you already know. Those prompts can only confirm what you already believe, and the point of this exercise is to find the names you didn't expect.
3. Run the prompts and record what you see
Run each prompt on the AI tools your buyers are likely to use. For a first pass, a simple routine works well: check each prompt on two relevant platforms, and repeat the check on three different days. That's a practical screening routine for finding candidates and not a statistically tested sample size.
The repeats matter because answers change. A 2026 preprint on repeated AI answers ran the same prompts again and again across four engines and found that the sets of brands and cited sources shifted a lot from one day to the next. For brand sets, the overlap between consecutive days ranged from about 45% to 59%. A separate study by SparkToro and Gumshoe.ai found that repeated recommendation lists were rarely identical. So one answer is a lead to follow up, and it isn't a ranking.
Context matters too. Google notes that search results can vary with things like personalization, language, location, and device, and AI Mode may use your past activity when personalization is on. So write down the region or language of each run, and don't assume any single setting removes the variation.
For every observation, record the exact prompt, the platform, the date, and the answer text if you can keep it. Then note each brand named and each source cited, and whether the brand was recommended, compared, or only mentioned in passing. Keeping the text lets a reviewer tell a real recommendation from a throwaway mention.
You're done when every candidate has at least one captured answer behind it. A candidate you saw only once should be marked as seen once.
A common mistake is picking competitors from a single answer, or copying a generated list as if it were a stable ranking.
4. Separate brands from the sources AI cites
Now sort everything you collected into four roles before anything goes on the shortlist:
- AI-visible brand competitor: a brand named as an option, alternative, or comparison in relevant answers.
- Buyer-comparison competitor: a brand the prompt or answer frames as a direct alternative to you, whether or not your sales team lists it.
- Cited source: a website, review publisher, marketplace, or community that the answer cites as support.
- Known rival, not observed: a competitor your team knows about that didn't show up in your sample.
AI can name one brand while citing a third-party review site for it. That review site may shape the answer, but it isn't selling a competing product. General sources such as Wikipedia, Reddit, and social platforms often show up among citations, and they belong in a source register, not in your brand list.
An Ahrefs guide to AI search competitor analysis points to three kinds of candidates: brands mentioned alongside you, domains that get cited instead of your site, and brands buyers compare you with. That's a useful way to find names, and the AI mention tracking tools roundup shows which trackers surface them. It still helps to keep them in separate lists.
You're done when every candidate has an entity type and a role, and the brand shortlist contains only brands.
Common mistake: "Competitor" can mean a company competing for the sale, a brand that appears in the answer, or a site that earns citations. Those are three different questions. Keep them in separate columns so your AI share of voice competitors comparison doesn't quietly mix brands with publishers.
5. Normalize names and judge relevance
Create a candidate register, which is just a sheet with one row per brand. Give each row a canonical brand name, known spelling variants, the company or product it refers to, its website domain, its role from step 4, and a short evidence note with the prompt, platform, and date.
Brand names are messier than they look. The preprint above points out that detection can miss abbreviations and paraphrases, and can also produce false matches when a common word happens to overlap with a brand name. So keep your own list of canonical names and aliases, and review likely duplicates by hand. If a company has a product name and a company name, decide whether they count as one competitor for this comparison, and write the decision down.
Then ask a few questions about each candidate:
- Does it solve the same problem for the same buyer you defined in step 1?
- Does it show up in more than one prompt, platform, or day?
- Was it recommended, compared, or named as an alternative, or only mentioned in passing?
- Is the name unambiguous, or could it be an alias or a false match?
Showing up across several different buyer questions is a helpful signal, though it doesn't prove a company matters in the market. Please don't invent a numeric score and present it as a validated ranking method, because nothing here supports one.
You're done when each brand on the shortlist has a canonical name, a short reason, and at least one evidence record, and any unclear identities are flagged for review.
Be careful not to count "Acme," "Acme Co.," and an Acme product as three rivals before you've checked whether they're the same company.
6. Cap the active list at a number you can act on
Start with five active brand competitors. There's no established ideal number here. One published guide suggests three to five, and another suggests roughly eight to fifteen once brands, publishers, marketplaces, and authority sites are all combined, which is the mix this guide tells you to keep apart. Whatever the count, the point is a group small enough to support a clean competitor benchmark AI visibility view. So five is a workload default, not an industry standard. If five is more than your team can review, start with three. You can add more later when you have the capacity to act on them.
To fill the slots, go in this order:
- Relevant brands recommended or named as alternatives across several of your buyer prompts.
- Brands that buyers directly compare you with.
- Relevant brands seen only once, but in a high-intent prompt.
A known sales rival that never appeared in your answers doesn't get an automatic slot. Move it to the "not observed" watchlist, and remember that this only means it wasn't seen in your sample, not that it never appears. Publishers and other source domains stay in their own register, and shouldn't be used to fill the remaining slots.
The reason for a cap is practical. A long "everything that might be a competitor" list spreads your attention thin, and it makes any competitor benchmark AI visibility comparison harder to trust, because the result depends on who is in the group.
Pro tip: If you only find four relevant brands, stop at four. Adding a weak fifth just to reach the number makes the list less useful.
You're done when the active list is small enough to review in one sitting, every name has a one-line reason, and your source domains and unobserved rivals are saved in separate lists.
7. Set up the list in your tracking tool and check it
Now put the shortlist into whatever you use for competitors AI search tracking, along with your prompt categories and the engines you plan to watch. Confirm how your own brand is named, the competitor list, how often collection runs, and which engines your plan covers.
In DeepSmith, the AI Visibility settings hold what counts as your brand and who your competitors are. The platform checks your chosen prompts on a schedule, saves the answers, and tracks mentions and citations separately. If you want to see how other tools handle that split, the AI citation tracking tools comparison lays it out. The overview shows a competitor leaderboard and the sources AI cites most, which is a quick way to confirm your shortlist went in the way you meant. DeepSmith doesn't find your competitors for you, so the list you built in the earlier steps is what you enter.
Check your plan for engine coverage too. Pro covers ChatGPT, Grow adds Perplexity, Scale adds Gemini, and Enterprise covers all ten supported engines. If you're still choosing a tracker, the best AI visibility tools roundup is a good starting point. The Profound vs Peec vs Ahrefs Brand Radar comparison shows how three trackers handle competitor views. Our AI share-of-voice tools roundup covers the share side of it. DeepSmith vs Rankscale is one specific head-to-head. It also helps to check how each tool's competitive view is set up before you commit.
One note about Content Map. It maps your website and competitor websites onto shared topics, and it has no cap on how many competitor sites you can add. That is a separate feature from the AI Visibility competitor list, so don't add every Content Map site to your tracking list just because there's room.
You're done when the configured list matches your saved shortlist, the prompts reflect your buyer questions, and you've checked the engines and schedule on your plan.
8. Keep the list stable and log every change
Save a dated copy of your active list and the reason for each name. Come back to it when you enter a new market, launch a relevant product, or notice your buyers changing. When you add or remove a competitor, write down the date and why, and treat the numbers after that as a new comparison period.
This matters more than it sounds, and it is easy to overlook when you're busy. The denominator for share of voice is whoever you include, so adding or dropping a brand can change the percentage even when the AI answers haven't moved at all. Different sources describe different ways of calculating AI share of voice competitors figures, so the safe rule is that numbers from differently defined groups shouldn't be compared as if they were the same. If you want a cleaner read on results, see how competitive benchmarking is handled in other tools.
You're done when a future reviewer can see which competitors were included in a period, what changed, when, and why.
The mistake to watch for is quietly changing names or membership and then crediting a shift in share to an AI engine or the market.
A quick worked example
Say a team sells project management software to distributed agencies. They draft about 20 to 25 prompts about agency workflows, client collaboration, integrations, and choosing software, then run them on two platforms across three days. They find four other software brands named as alternatives. They also see a software review publication and a community forum cited in many answers.
The four brands go into the candidate register. The publication and the forum go into the source watchlist. A fifth company from the sales list never appeared, so it stays on the "known but not observed" list. The team starts with four tracked brands, and would add a fifth only if another relevant brand turns up in the evidence. This is an illustration of the method and isn't real customer data.
What to do next
Save your list with today's date, write down the reasons behind each name, run your prompts on a regular schedule, and only revisit the group when the market or your evidence changes. Once your tracking is live, you'll have a steady base to work from.
If you'd like to try this with your own prompts and competitors in one place, you can start a free DeepSmith trial and set up your list in the AI Visibility settings.
Related guides
- Oltre Alternatives: 8 Best AI Visibility and Content Tools for 2026: Compares eight Oltre substitutes (DeepSmith, Frase, Writesonic, Surfer, Profound, Athena, Goodie, Semrush) for buyers who want AI plus Google visibility tracking paired with content production, explaining which tools.
- SE Visible Alternatives: 7 Best AI Visibility Trackers for 2026: Compares seven substitutes for SE Ranking's SE Visible: DeepSmith, AthenaHQ, LLMrefs, RankScale, OtterlyAI, Peec AI and Profound.
- Siftly Alternatives: 8 Best Tools to Track How AI Describes Your Brand in 2026: For teams watching how AI models describe their brand, this ranks eight Siftly alternatives (DeepSmith, Profound, Peec AI, OtterlyAI, LLMrefs, Scrunch, AthenaHQ, Rankscale) on multi-engine tracking, prompt-level sentiment.
- Surfer AI Tracker Alternatives: 7 Best AI Visibility Tools for 2026: Scoped to Surfer's AI Tracker feature rather than its content editor, this details Surfer's five tracked surfaces and plan tiers, then compares seven alternatives (DeepSmith, AthenaHQ, Profound, LLMrefs.
- Troovue Alternatives: 8 Best AI Citability Diagnostic Tools for 2026: Explains why Troovue's 0 to 100 Visibility Eligibility Index is a readiness condition, not a citation outcome, then ranks eight alternatives (DeepSmith, Profound, Peec AI, Scrunch, AthenaHQ, Otterly.


