How DeepSeek recommends
Why you don't come up on DeepSeek
Mador has read 508 answers from DeepSeek. Of 87 businesses measured, 83.9% never show up in DeepSeek's answers. DeepSeek answers with what it has learned, without searching the web: that's why it names whoever is already known online.
Every assistant reads the web its own way
Each assistant reads a different number of sites before answering, and names whoever it finds there. If your business isn't on the sites it reads, for that assistant it doesn't exist.
- ChatGPT7,284 answerssources analysed in 6,914Sites4.3Says sites81%Portals12%
- Claude4,628 answerssources analysed in 4,258Sites5.3Says sites48%Portals24%
- DeepSeek508 answersDoesn't search the web: it answers from memory.
- Gemini4,701 answerssources analysed in 4,331Sites6.1Says sites96%Portals8%
- GLM498 answerssources analysed in 378Sites9.4Says sites100%Portals27%
- Grok517 answerssources analysed in 398Sites6.0Says sites100%Portals10%
- Kimi403 answersIt searches, but hands back no address.
- Meta AI560 answerssources analysed in 440Sites1.6Says sites80%Portals18%
- Mistral415 answerssources analysed in 295Sites4.6Says sites98%Portals8%
- Perplexity6,840 answerssources analysed in 6,470Sites11.9Says sites100%Portals25%
- Qwen538 answersIt searches, but hands back no address.
The 11 platforms in this table, out of the 14 Mador measures, do not share one measurement window: the shortest belongs to Claude, 75 days; the longest to Perplexity, 106 days. The numbers in each row hold for its own window, not for a period they all have in common.
Mador measures Bing (3) too, but under 50 answers the numbers move with the next scan, so they don't appear here: measured, not yet enough.
The answers are the ones each assistant really gave, counted once, including those from Mador's studies and reports. Figures updated on November 12, 2026.
DeepSeek doesn't search the web: it answers with what it already knows.
How it recommends
DeepSeek doesn't search the web and names few businesses, three or four per answer on average: it takes them from what it has learned. To get in, what counts is being known online for a while, and the platform tells you when you start to show up.
How much it's used.
Every assistant has its own audience, and its users trust the names it gives. If DeepSeek doesn't name you, those people don't find you.
3%
of web visits
Similarweb · September 2026
less than 1%
of clicks to websites
Statcounter · October 2026
not measured by this source
Sensor Tower · May 2026
It answers from memory.
DeepSeek doesn't search the web while it answers: it uses what it learned beforehand. That's why it changes slowly, and names whoever was already known online. The good news is that the presence you build today can make it into its future versions, and the platform tells you when you start to show up.
No list of sites.
DeepSeek doesn't check the web while answering, so there's no list here. To show up in its answers, what counts is a steady presence online, on the same sites the other assistants read.
You come up on one and not the other
When a business gets named, in two cases out of three only one of the two assistants does it. Whoever shows up on DeepSeek often doesn't show up on ChatGPT, and the other way round. Your customers don't all use the same one: that's why the platform measures them all, and tells you where you're missing.
Same practice, two different answers.
nome coperto, a dentist in Turin. In the real measurement the platform asked ChatGPT and Perplexity, the two assistants in the free scan, the same five questions. Perplexity named it in three, ChatGPT in none.
The questions in this reading
- 1.Qual è il dentista migliore a Torino?
- 2.Dove mi conviene andare dal dentista qui a Torino?
- 3.Quale studio dentistico preferire a Torino se ho paura?
- 4.Mi consigliate un dentista bravo e onesto a Torino?
- 5.C'è un dentista consigliato in centro a Torino?
Answer by answer
If this were your business and you only looked at ChatGPT, you'd think you didn't exist. On Perplexity, with the same questions, it comes up three times out of five. The good news is that in a few minutes you can see where you really stand, on both.
Frequently asked questions
Why doesn't DeepSeek name me?
Because it answers with what it has learned, and your name isn't there yet, or barely. 83.9% of the businesses measured are never named. You get in by having a steady presence online, on the sites the other assistants read too.
How do I get DeepSeek to name me?
By having a steady presence online, on the same sites the other assistants read. The free scan shows you them in a few minutes, then Mador writes the pages you're missing and lets you know when something changes.
Does DeepSeek say where it gets its answers?
No, because it doesn't search the web: it answers with what it knows. The platform looks at the names inside the answer, which is what matters to you.
If I come up on ChatGPT, do I come up on DeepSeek too?
Not necessarily. When ChatGPT names you, DeepSeek does too in less than half of cases. They choose differently, which is why it pays to watch them all.
What data are these figures based on?
We count the answers DeepSeek really gave from August 27, 2026 to November 9, 2026: 508. Each answer counts once, when the assistant gave it; an answer read again from our archive doesn't count twice. They include the answers from Mador's studies and reports, which are real answers. The businesses are 87, in 9 areas and 13 categories: only real businesses, without the subjects of studies and reports and without our own test scans. Figures updated on November 12, 2026.
Let's see whether you come up, and who comes up instead of you
In a few minutes you'll know whether ChatGPT and Perplexity name you, and which names they give instead of yours. With a plan you measure DeepSeek too. It's free, and your email gets you into the platform.