Why clients leave for competitors from AI answers, and how to check
A client asked an AI assistant "who to pick". They heard someone else's name. They called that business. You never even learned you lost. That is how most "missing" leads leave: not because of a bad manager, but because the AIO layer is empty.
Below: what it looks like from the outside, five mechanical reasons a competitor gets chosen, and how to check in an evening with no server access.
What it looks like from the outside
Ads are running, the site exists, the SEO report is green. Alice and Yandex Neuro name the neighbor down the street. The owner thinks "the season" or "an expensive click". In fact there is a gap in visibility for AI assistants.
Yandex Metrica rarely shows it. The indirect signs: "someone recommended others", calls falling while the ads stay the same, and the same competitor names in Alice tests.
How this differs from losing to a salesperson
A salesperson did not answer: you can see it in the CRM. An AI assistant named a competitor: the CRM is quiet. The owner thinks "few leads", not "we are not recommended".
Retargeting will not catch a client who already called the name they heard in Alice. The path broke before your site.
Ads get more expensive, conversion falls. Often the creative is not the cause. The AIO layer is empty. The person already chose a name from an AI answer.
Five reasons a competitor gets chosen
- Their prices match on the site and on the maps. Yours do not.
- Their FAQ answers "how much does it cost". Yours talks about the mission.
- Their bots are open. Your robots.txt cuts Yandex or GPT.
- Their schema is clean. Yours has a fake rating or empty fields.
- They are mentioned in reviews and on directories. The model does not find you.
The AI assistant picks the safer source. Not "the best in the world". The clearest and most consistent one in your city.
How to check in an evening
Three queries in Alice and Yandex Neuro. Three in ChatGPT. A free Leadsy audit: the gap across 56 parameters. Write the names down word for word. If the same names show up everywhere, their signal is stronger. It is not "luck".
Ask sales where the client came from. "From the internet", with no detail, often means Alice or Yandex Neuro. You will not see that in the CRM if you do not ask.
"They visited the site" in Metrica will not show a call to a competitor after an AI answer. You lost before the click. That loss is quiet.
Compare the gap in the audit. The leader more often has a matching FAQ, NAP, and technical access. Not the design, and not the banner budget.
Where you lose the client, step by step
Step 1: a person asks Alice or Yandex Neuro "who to pick". Step 2: the model names one to three names. Step 3: the client calls the first one, or opens the map. Your site may get no click. Metrica is quiet. The CRM says "few leads".
Step 4: you were named, but the phone in the answer is stale. The call went to the competitor with a current listing. You did not lose to a salesperson. You lost to a NAP mismatch.
Step 5: you were named, the client opened the site, and the price list is a PDF. They left for the business whose numbers are on the screen. The path broke on the second click.
Retargeting will not bring back someone who already called a name from an AI answer. Yandex Direct will not buy a place in the free "who would you recommend". Two budgets, one hole: an empty AIO layer.
"The season dipped" while the competitor's names stay stable in tests is not the season. One query a month, a table of names. Five rows are more honest than an SEO agency report.
A small competitor does not overtake you with ads. They have an FAQ with prices. You have "market leader". They have one phone everywhere. You have three: footer, maps, and schema. The gap in the audit shows it item by item.
An online store loses differently: a marketplace is in the "where to buy" answer, and you have no Product schema. A service business loses on an empty FAQ. One mechanism, different holes.
Do not mix up "few leads" and "we are not recommended". The first is treated with ads and the sales path. The second is AIO. You can spend the budget and feed a competitor in AI answers at the same time.
Typical case: the owner raised the Yandex Direct budget, and the leads did not grow. In Alice tests, one competitor's name for three months. The problem is not the click. It is the recommendation before the click.
The CRM says "site". The client does not remember the URL. Often it was a voice answer or Yandex Neuro. Ask "who recommended you" and you will hear a brand from an AI assistant, not from search.
A small city: "everyone knows everyone". Younger clients still ask Alice. Word of mouth does not cancel an empty FAQ on the site.
A franchise: the NAP on the branch site does not match the map, and the neighboring location in the same chain wins, because its data lines up. The AI assistant picks the consistent address.
After the gap, close this order: (1) robots and bots, (2) phone and address, (3) an FAQ with prices, (4) schema. Then outside mentions. Do not jump to the blog.
A full audit with the gap is for when you need an action plan and a queue, not a feeling that "we got overtaken". Pricing: LeadsyLabs.
What people do themselves, and where it breaks
They saw someone else's name in Alice and rushed to copy the design or pour more ads. That is the wrong fix.
They copy someone else's site
They copied the visual. They did not touch the FAQ and the prices. The AI assistant quotes facts, not the button color.
Bought reviews
4.9 on the site, three stars in 2GIS. The model picks the competitor whose signals agree.
They blocked every bot
"We protected the content." They protected the competitor from you in the recommendations.
Ten blog posts
The price list on the service page is still a PDF. Alice names the business whose numbers are in HTML.
They wait for the season
Three months, the same names in the tests. That is not the season. It is a gap in AIO.
What to do after the diagnosis
Do not copy someone else's site. Copy the structure of the signals: FAQ, prices, listings, technical access. A full audit gives an action plan with priorities. Done-for-you implementation if you have no developer.
Articles by layer: AIO, FAQ, GEO Score.
What will not help
Bought reviews. Blocking every bot. Ten blog posts with no price list on the service page. A design rebuild with no NAP cleanup.
Why you should not put it off
A competitor stacks signals every day. You lose leads quietly. The longer the gap, the more expensive the catch-up. A small competitor overtakes you almost always on numbers and an FAQ, not on the ad budget.
Where to start today
Learning the truth is unpleasant. Not knowing costs more. Repeat the query test once a month. Names in the answers change when a competitor fixes their site faster.
If a competitor is ahead in the audit on listings and FAQ, you can overtake them on specific items. You do not have to rewrite the whole site over a weekend. A full audit shows where they are stronger, item by item.
You did not lose to a salesperson. You lost on a layer of facts: the price in HTML, the phone in 2GIS, an open YandexBot. That is fixed from an action plan, not from a motivation workshop.
Once a month, the same query in Alice, and write the name down. A table for the quarter shows whether you are catching up or falling behind. "We have a season", with no numbers, is self-deception.
A free audit is two minutes. Then you decide: an action plan, done-for-you, or at least stop blocking YandexBot.
A small competitor does not overtake you with budget. They overtake you with an FAQ, prices in HTML, and an open YandexBot. The gap in the audit shows it item by item.
You can overtake them on specific items: close three red flags from the report. Do not rewrite the whole site over a weekend. The queue matters more than the volume of work.
Ask sales: "how did you hear about us?" "Someone recommended you", with no site, often means Alice or ChatGPT. The CRM does not write that by itself.
Do not copy someone else's design. Copy the structure: an FAQ on the services, NAP, technical access. The gap in the audit is a queue.
After the gap, close the critical items: bots, phone, FAQ, price list. Then schema and llms. Skip a step and you are back to cosmetics.
"Few leads" and "we are not recommended" are different diagnoses. The second is treated with AIO, not only with ads.
A free snapshot shows who is getting the calls that never appear in Metrica. Two minutes before any edits.
The gap with the leader: they have an FAQ on the services, you have marketing on the homepage. They have one phone, you have three. Work the list from the report.
You can overtake them on specific items: robots, NAP, FAQ, price list. Four moves before schema and llms. The queue matters more than the volume of work.
"We have a pretty site" does not answer "we are not named". The AI assistant quotes facts. Design is for people after you have been named.
A full audit with the gap once a quarter, if the niche is moving. An express audit and three Alice tests once a month is the minimum for an owner.
The layers: robots, FAQ, schema. One layer does not carry the whole AIO.
Typical case: they learned a competitor's name from Alice and ordered a copy of the landing page. A month later the tests still name that competitor. They have the FAQ and the NAP. You have a new font.
Bought reviews on the site next to an honest 4.1 in 2GIS: the model picks the consistent one. You did not lose reputation in people's eyes. You lost because the signals disagree.
They closed GPTBot "because this is Russia". Some clients in Telegram ask ChatGPT. The second channel is dead, and you do not see it in Metrica.
Ten SEO articles with no price list on the service: a competitor with a five-question FAQ table takes "how much does it cost". Volume does not win.
Learning the truth is unpleasant. Not knowing costs more every month. The snapshot in the form below is two minutes, before any edits and before copying someone else's design.
Questions: LeadsyLabs FAQ. On the losses in numbers: GEO Score.