Why AI assistants recommend your competitor (and what to do about it)

A practical guide from AEO Pulse · Updated June 2026 · ~8 minute read

It stings. A customer mentions they asked ChatGPT for "the best roofer around here" and it named your competitor — twice, with reasons — and never mentioned you. Before you take it personally: an AI assistant's recommendation is not a judgment of your work. It's a summary of the evidence about you that exists on the web, assembled in a few seconds by software with limited patience.

That's actually good news, because evidence gaps are fixable. This guide explains the mechanics — why one business gets named and another doesn't — and then gives you a prioritized fix list.

First, why this is worth fixing at all

Two published data points frame the stakes. Ahrefs studied 300,000 keywords and found that when Google shows an AI Overview, the click-through rate of the #1 organic result was 34.5% lower — and their December 2025 follow-up put the reduction at 58%. Meanwhile Otterly.AI's research reports 15% of total website traffic now arrives via AI agents and bots, with ChatGPT alone accounting for 56% of AI referral traffic. The AI's short answer is increasingly the whole decision. When it names two businesses and you're not one of them, you don't lose the click — you lose the existence.

How an assistant picks who to recommend

When someone asks a web-search-enabled assistant "best HVAC company in Mesa", it typically searches the web like a fast, impatient researcher: it pulls a handful of sources — review platforms, local directories, "best HVAC in Mesa"-style articles, business websites — and names the businesses that show up repeatedly, recently, and positively across them, usually with a one-line justification ("4.8 stars across 500+ reviews", "family-owned, 20 years in business").

So when your competitor gets named and you don't, the cause is almost always one or more of these six:

Reason 1: They have more — and more recent — review evidence

Review counts, ratings, and recency are the justification assistants reach for most often, because they're the most quotable evidence available. A competitor with 400 reviews, dozens from this year, gives the assistant a safe, citable recommendation. Your 23 reviews — even if they're glowing — are thinner evidence, and thin evidence gets skipped when the answer only has room for three names.

Reason 2: They're on the lists the assistant reads

Ask an assistant for "the best [trade] in [city]" and watch its sources: very often it leans on a few existing "best of" roundups — local media, trade sites, directories. Businesses on those lists get recommended; businesses absent from them effectively don't exist for that query pattern. Your competitor may simply have been included in two listicles three years ago and has been harvesting AI recommendations ever since.

Reason 3: Their identity is consistent; yours is fragmented

Assistants connect mentions across sources to build confidence. If you're "Smith & Sons Roofing" on Google, "Smith Roofing LLC" on Yelp, and "smithroofingtx.com" never states a city — the assistant can't be sure those are the same business, and uncertainty reads as risk. Inconsistent name, address, phone, and category data quietly disqualifies businesses from answers.

Reason 4: Their website answers questions; yours is a brochure

An assistant skimming your competitor's site finds "We repair and replace residential roofs across Mesa, Gilbert, and Chandler. Free inspections. Licensed ROC #123456." — quotable, specific, geographic. Yours says "Quality You Can Trust" over a stock photo. One of these gives an AI something to say; the other doesn't.

Reason 5: You're being confused with someone else

Similar business names across cities are a real failure mode: the assistant describes a different "Lakeside Plumbing" three states away, or blends two businesses' details. If your name is common, the disambiguating evidence (consistent city mentions, distinct branding, schema markup) matters double.

Reason 6: Plain luck of the draw — sampled once

AI answers are probabilistic: the same question can produce different names on different runs. If a customer's single ChatGPT answer skipped you, that one run isn't proof of a pattern. Measure properly — multiple runs, multiple days — before (and after) you fix anything. Our DIY audit guide shows exactly how, or the free check does a structured version for you.

The fix list, in priority order

1. Measure your baseline (this week, ~15 minutes). Run the standard queries — "best [category] in [city]", "who do you recommend for [category] near [city]", "[business] reviews", "tell me about [business]" — in fresh AI conversations, and record who got named and which sources were cited. Those cited sources are your literal to-do list.
2. Build review momentum (start now, never stop). Make the ask part of job completion: every happy customer, same-day, with a direct link. Recency matters as much as volume — a steady trickle of genuine current reviews out-evidences an old pile. (Genuine only: fake reviews risk platform penalties and poison the well.)
3. Fix identity consistency (one afternoon). One exact business name, address, phone, category, and website across Google Business Profile, Yelp, Bing Places, and your trade's main directories. Make your website state your city and service area in plain text on the homepage.
4. Pursue the lists that assistants cite (ongoing). From step 1 you know which "best of" pages drive answers in your market. Many accept submissions, review-based inclusion, or simple outreach. Legitimate inclusion on two or three of those pages directly changes the inputs assistants read.
5. Make your site quotable (a weekend). Add plain-text answers to the questions customers actually ask: what you do, where, pricing approach, response time, licenses/credentials, FAQ. Add LocalBusiness and FAQPage schema markup so crawlers parse it cleanly.
6. Re-measure and iterate (weekly or monthly). Same queries, same method, logged with dates. AI answers move when the web moves; the log tells you what worked. This loop — measure, fix evidence, re-measure — is the entire discipline. (It's also exactly what our founder plan automates, if you'd rather not run it by hand.)

What doesn't work (save your money)

• Buying placement. There is no ad product that buys your way into ChatGPT, Claude, or Gemini organic answer text for local recommendations. Anyone selling "guaranteed AI placement" is selling something they don't control.
• Keyword-stuffing your site for AI. Assistants summarize meaning, not keyword density. "Best plumber Austin best plumbing Austin TX plumber" reads as spam to humans and models alike.
• Fake reviews and astroturfed listicles. Beyond the ethics: review platforms actively police this, and a platform penalty destroys the exact evidence base you're trying to build.
• One-time fixes with no measurement. If you don't re-run the queries, you're optimizing blind. Measurement is the cheap part — don't skip it.

An honest note about guarantees

Nobody — not us, not a $499/month enterprise platform, not an agency — controls what AI assistants say. The honest claim is narrower and still valuable: you can know what they say, improve the evidence they rely on, and verify whether it moved. Businesses that run that loop have a structural advantage over the majority who have never once checked.

FAQ

Why does ChatGPT recommend my competitor instead of me?

Usually one or more of: they have more recent review evidence, they're on the "best of" lists assistants cite, their business identity is more consistent across the web, or their website states plainly what they do and where. It's evidence, not verdict — and evidence is fixable.

Can I pay to be recommended by AI assistants?

No. There's no placement product for organic AI answer text. You influence the inputs (reviews, listings, lists, website), not the output directly.

How long until fixes show up in AI answers?

No fixed timeline — assistants reflect the web once the pages they read change. Re-measure weekly or monthly and track movement instead of expecting a scheduled result.

What's the single highest-impact fix?

For most local businesses: recent review volume on the platforms assistants cite, with inclusion on the specific "best of" pages they reference a close second. Your own baseline check will show which gap is yours.

Related guides: What is AEO? AI search optimization in plain English · How to check what ChatGPT says about your business (free methods)