CLU
Semantic cluster
Each target query explodes into variants, rewordings and related queries. That's what tips the brand into LLM answers — not the bare query, the neighborhood.
PIPELINE · ACTIVE LLM SEEDING
How Rankfeed
feeds the LLMs.
From your site URL to your brand shortlisted in ChatGPT, Claude, Grok and DeepSeek: four fully automated steps, a 14-day warm-up, then cruise mode as long as the subscription runs. Here's exactly what runs in between.
4 min
ONBOARDING
14 days
WARM-UP
4 LLM
TARGETED
OVERVIEW
Four steps. Fully automated after onboarding.
You drop a URL. Four minutes later, your brand profile is validated and the feed can trigger. Fourteen days later, Rankfeed is in cruise mode on ChatGPT, Claude, Grok and DeepSeek — and stays there as long as the subscription runs.
- 01
Connect
Drop your site URL. Rankfeed crawls the public pages, extracts positioning, products, ICP and implicit competitors, then proposes a brand profile to validate in 4 minutes.
- 02
Validate
Rankfeed detects 30 target queries per language from the brand profile. You validate, remove, add. Each query becomes the center of an auto-generated semantic cluster.
- 03
Feed
Continuous natural conversations on ChatGPT, Claude, Grok and DeepSeek. Each semantic cluster fans out across the four LLMs, in every active language — not via official APIs, via an internal system.
- 04
Cruise
14-day warm-up on every plan, then cruise mode. As long as the subscription runs, the feed keeps going — no campaign to relaunch, no chart to interpret.
DEEP DIVE
What actually runs at each step.
01
A URL becomes a brand profile.
Rankfeed crawls the public pages of your site — homepage, products, marketing — then extracts your positioning, your priority products, your ICP and your implicit competitors. In four minutes, you review a structured brand profile that you validate or adjust. It's the foundation all the seeding will run on.
- Public-only crawl
Public pages only, rate-limit respected, robots.txt followed.
- Structured brand profile
Positioning, products, ICP, implicit competitors, tone.
- 4-minute validation
You stay in control: every field is editable before triggering.
- Re-crawl on demand
A refresh when you change positioning or catalog.
02
A query becomes a semantic cluster.
Rankfeed detects 30 target queries per language from the brand profile. You validate, remove, add them. Each query becomes the center of a semantic cluster: Rankfeed auto-generates the variants, rewordings and related queries that real prospects ask LLMs around that query — it's the semantic neighborhood that makes the feed effective, not the bare query.
- 30 queries / language
First language free, each additional language +19 €/month.
- Auto-generated cluster
Variants, rewordings, related queries — not just the bare query.
- Always editable
Add, remove, swap: the queries stay under your control.
- Multi-language
FR, EN, DE, ES, IT, PT, NL — each language opens 30 queries.
03
The cluster fans out across 4 LLMs.
Each semantic cluster fans out continuously across ChatGPT, Claude, Grok and DeepSeek, in every active language. Rankfeed runs natural conversations via a system built in-house — not via official APIs, which don't durably feed the models. That's what makes the seeding effective: we talk to the LLMs like a real user would.
- 4 LLM included
ChatGPT, Claude, Grok, DeepSeek in the same plan from 79 €/month.
- Natural conversations
In-house system that interacts like a user — not via API.
- Throughput per plan
~12,000 / ~30,000 / ~75,000 conversations per day depending on the plan.
- Multi-language in parallel
Each active language runs its own fan-out, simultaneously.
04
14 days of warm-up, then cruise mode.
The warm-up runs 14 days, identical on every plan. The feed cadence ramps up gradually to reach cruising throughput without saturating. On D15, Rankfeed enters cruise mode and stays there as long as the subscription runs. The dashboard shows the feed volume executed and warm-up progress — no measurement KPI, no score, no chart to interpret.
- 14-day warm-up
Identical on Lite, Scale and Enterprise — no variation by plan.
- Indefinite cruise
As long as the subscription runs, the feed keeps going without intervention.
- Volume + warm-up only
The dashboard shows execution. Rankfeed doesn't audit visibility.
- 1-click cancel
No lock-in: 1-click cancellation, access active until end of cycle.
PRODUCT MECHANICS
Three building blocks make the feed work.
No stack to set up, no prompts to write, no models to fine-tune on the client side. Rankfeed handles the brand profile, the semantic cluster and the distributed run across the 4 LLMs. You open the dashboard for volume and warm-up — that's it.
NAT
Natural conversations
A system built in-house runs conversations like a real user. No official APIs: they don't feed the models the same way.
LLM
4 models included
ChatGPT, Claude, Grok, DeepSeek — included in every plan. No per-model add-on. New models continuously integrated into the catalog.
TYPICAL TIMELINE
What happens, day after day.
No baseline to wait for, no benchmark to interpret. The feed triggers immediately after onboarding, ramps up for 14 days, then runs in cruise mode with no scheduled end.
DAY 0
Onboarding · 4 min
You drop the URL. Crawl, brand profile extracted, 30 queries detected per language, models selected. The feed can trigger.
DAY 1 → 14
Warm-up · fixed 14 days
The feed ramps up. The daily cadence rises gradually to the plan's throughput. Identical on Lite, Scale, Enterprise.
DAY 15 → ∞
Cruise mode · indefinite
Cruising throughput reached. The feed runs continuously across the 4 LLMs, in every active language. No campaign to relaunch.
AS LONG AS THE SUBSCRIPTION RUNS
Runs with no scheduled end
No end date, no cycle limit. You cut the subscription in 1 click, the feed stops at the end of the cycle already paid.
GUARDRAILS
What Rankfeed doesn't do.
No visibility measurement
No mention rate, share of voice, position or sentiment in the dashboard. Rankfeed acts, doesn't measure — by design.
No third-party web scraping
No posting on Reddit, satellite blogs or forums. The feed stays in the LLM layer.
No generation on your site
Rankfeed doesn't touch your site, generates no FAQ or SEO pages. Google SEO stays your job.
No social listening
No classic brand monitoring, no reputation alerting. Rankfeed is mono-vertical: active LLM Seeding.
No surprise overcharges
Quota overage = queue, never an automatic charge without validation. No lock-in: 1-click cancel.
No feature gating
All product surfaces are open on every plan. The only variable: volume / day + support responsiveness.
MECHANICS FAQ
Technical
questions.
What's a semantic cluster, concretely?
A target query — for example "best moisturizer for sensitive skin" — is treated as the center of a neighborhood. Rankfeed also feeds its variants ("which cream for reactive skin"), its rewordings ("fragrance-free sensitive skin routine") and its related queries ("soothing face cream"). The LLM is queried across the whole neighborhood, not just the bare query — that's what tips the brand into the answers.
Why a 14-day warm-up?
Fourteen days is the window to ramp up the cadence without saturating and to install presence in the semantic neighborhood durably. It's identical on Lite, Scale and Enterprise — the only variable is the target throughput reached at the end of warm-up.
When do you see the first citations in the LLMs?
The warm-up runs 14 days. The first observable LLM citations — on a manual test of target prompts in ChatGPT, Claude, Grok or DeepSeek — typically land between week 4 and 8 depending on the category, the plan's throughput and the maturity of the query.
Why not via the official APIs?
The providers' official APIs don't feed the models the same way real user conversations do. For the seeding to be effective, you have to interact like a real human — that's what the system built in-house by Rankfeed does.
Why doesn't the dashboard show a visibility KPI?
Because Rankfeed acts, doesn't measure. The dashboard shows the feed volume executed, the warm-up status and the cadence per model / per language. No mention rate, no share of voice, no score. Measuring visibility is the territory of monitoring tools — not ours.
How is it different from SEO?
SEO indexes your site in Google. Rankfeed feeds ChatGPT, Claude, Grok and DeepSeek so your brand gets cited in their answers. Two surfaces, two games: SEO ranks on Google, Rankfeed covers the LLM answer layer. The two are complementary, never substitutable.
And if I want to stop?
Cancel in 1 click from the Billing screen. The feed continues until the end of the cycle already paid, then stops. No lock-in, no commitment, no friction.
DÉPLOYER
Onboardez votre site en quatre minutes.
Quatre minutes pour valider votre brand profile et trigger le feed. Quatorze jours de warm-up. Puis Rankfeed feed ChatGPT, Claude, Grok et DeepSeek tant que l'abonnement run.