How to Know if a Competitor Is Outranking You in AI Answers

How do I know if a competitor is outranking me in AI answers?

Test it directly: pick 8–12 buying-intent prompts your customers would ask, run each through ChatGPT, Gemini, and Perplexity several times, and count how often each brand — yours and each competitor’s — gets named. A rival is outranking you when it’s named more consistently across those runs than you are.

The word “outranking” is doing something slippery here, so let’s be precise before you start counting. AI answers don’t have a stable rank. Ask the same assistant the same question twice and the list of businesses it names can change. So “outranking you in AI” doesn’t mean a competitor sits at position #2 while you sit at #5 — it means the competitor shows up more often across many prompts and runs than you do. Frequency is the signal. A single screenshot is not.

Why this matters for a service business

Your prospects are asking AI who to hire, and the answer names a short list. BrightLocal’s 2026 research found that 45% of consumers now use AI tools like ChatGPT, Gemini, or Perplexity to find local businesses — up from 6% a year earlier — with ChatGPT specifically used by 31% of them. When someone asks “who’s the best [your service] near me,” a handful of firms make the cut and the rest are invisible. If your competitor is one of the named few and you aren’t, you’re losing jobs you never even knew were in play.

The scale makes it worse to ignore. ChatGPT alone reached roughly 900 million weekly active users in early 2026, so the audience quietly comparing you to your rivals inside a chat window is enormous. You can’t see those conversations. But you can reproduce them — by asking the same questions your buyers ask and recording who gets recommended.

How to check whether AI recommends your competitors over you

This is a repeatable process, not a one-time look. Do it once to get a baseline, then again monthly so you can watch the trend move.

1. Write your real buyer prompts

List the 8–12 questions a prospect would actually type before hiring someone in your category. Mix intents: “best [service] in [city],” “who should I hire for [problem],” “[Competitor] vs [you],” “affordable [service] near me,” and “top-rated [service] for [use case].” These decision-stage prompts are where recommendations happen — not “what is [service],” which returns definitions, not a shortlist. Keep the wording natural; write how a customer talks, not how you’d write a headline.

2. Name your competitor set

Decide which rivals you’re measuring against — usually your 3–7 closest local or category competitors. Add any brand that keeps surfacing in AI answers even if you don’t consider them a direct rival; AI doesn’t know your rivalries, and a firm the model loves is beating you whether you like it or not. Write the names down. You’ll tally mentions for each of them alongside your own on every run.

3. Run each prompt across multiple engines

Ask every prompt in ChatGPT, Google Gemini, and Perplexity — and, if your audience uses them, Microsoft Copilot and Google’s AI Mode. Different engines pull from different places: some cite live web sources (grounded answers), while others answer from training data (memory). Run each prompt in a fresh chat with no prior context, so an earlier question doesn’t bias the next answer. Copy each response somewhere you can review it — a spreadsheet works fine.

4. Run each prompt several times — this is the whole game

Because these models are probabilistic, the same prompt returns different brand lists on different runs. One study found ChatGPT gave consistent answers only about 73% of the time when asked the identical question ten times over. So run each prompt at least three to five times per engine. A brand named in one run out of five is a maybe; a brand named in five out of five is a genuine recommendation. This repetition is exactly what separates a real reading from a lucky (or unlucky) screenshot.

5. Tally mentions, not positions

For each brand — yours and every competitor — count how many runs named it, then turn that into a rate: “mentioned in 12 of 20 runs = 60%.” Do this per engine and overall. Now you have an honest, comparable number. If a competitor’s mention rate is 80% and yours is 25% on the same prompts, that competitor is outranking you in AI, and you can say so with evidence instead of a hunch. Resist the urge to record “we came 2nd” — position swaps run to run and means little.

6. Read the citations, then close the gap

For grounded answers, click the sources the AI cited. This is the gold. If the model keeps citing a competitor’s blog post, a “best [service]” listicle, a Reddit thread, or a directory, that’s the specific page winning the mention — and the exact thing you can go earn a spot on. Common levers: publish clear, answer-first pages for those buyer questions, get listed on the roundups the AI trusts, keep your business details consistent everywhere, and build genuine reviews. Then re-run the process next month and check whether your mention rate moved.

Manual checking vs. a tool

The six steps above are free and you should do them at least once — nothing teaches you faster than watching a competitor get named while you don’t. The catch is consistency. Doing 12 prompts × 3 engines × 5 runs by hand is 180 chats to open, read, and tally, every single month, worded identically each time so the comparison stays fair. Most owners do it once, learn something useful, and never keep it up.

That gap is where tools help. Dedicated AI-visibility platforms run your prompts on a schedule, count mentions across engines and runs automatically, compute your share of voice against competitors, and chart the trend. Full disclosure: ScoutRival is our tool. It runs your buyer prompts through two engines — ChatGPT’s web-grounded answers and a Google Gemini “model-memory” probe — and reports how often you and each competitor are mentioned, your share of voice, sentiment, and the citations behind grounded answers. Checks are run on demand, so you re-check any day rather than waiting on a cron job, and every answer is measured through official, grounded access — never by scraping the ChatGPT app, because there’s no legitimate public API for the consumer product. If you’d rather compare a few options first, see our roundup of the best AI visibility tools; alternatives like Profound, Otterly.AI, and Peec AI cover similar ground at different scales.

Whether you go manual or paid, the honesty test is the same: a credible check reports how often you’re mentioned across many runs, not a single fixed “rank.” Any tool or screenshot promising a stable ChatGPT position is selling certainty that doesn’t exist.

“We assumed AI didn’t matter for a plumbing business. Then we ran ten prompts and a competitor two towns over got named in eight of them — we got named in one. That one number changed our whole content plan,” — Dana Whitfield, home-services owner (illustrative).

Common mistakes that give you a false reading

A few errors will make your check lie to you:

• Judging from one run. The single most common mistake. One answer is noise; you need the rate across several runs before you conclude anything.
• Treating position as rank. “We came third” is meaningless when the order reshuffles every time. Count presence, not placement.
• Testing definition prompts. “What is [service]” won’t recommend anyone. Only buying-intent prompts surface a shortlist.
• Ignoring the citations. The named brands tell you who wins; the cited sources tell you why — and the “why” is the only thing you can act on.
• Checking once and stopping. AI answers shift as models update and the web changes. A baseline with no follow-up can’t show whether you’re gaining or losing ground.

This work sits inside your broader competitor-watching habit. If you’re building that from scratch, start with our guide on how to monitor competitors online, and pair AI checks with the classics: tracking your competitors’ Google rankings and measuring share of voice across channels. For more ways to run this without a marketing team, browse our marketing automation guides.