AI overview

Reynt has AI woven throughout the console to speed up the repetitive parts of running your business — writing listing content, translating it, importing properties, and drafting website and support text. This page explains what AI can do, how to switch it on, and how to keep an eye on cost.

What AI can do

Import a listing — paste a URL, text, or PDF, or upload photos, and AI extracts the details, writes the content in all your languages, and matches amenities. See Create a listing.
Generate and translate listing content — write or translate individual fields (title, description, SEO, and more) across your languages. See Content generation and translation.
Label listing images — vision analysis produces descriptive labels for photos.
Generate website content — draft content sections for CMS pages and blog posts.
Bootstrap your catalog — generate tags and cancellation policies tailored to your business.
Suggest icons — when picking an icon for a catalog item (amenity, attribute, and the like), Suggest with AI proposes matching icons based on the item’s name.
Draft support and checkout text — reusable canned responses for the inbox, and “How did you hear about us?” referrer options.
Reply to guests — draft, or automatically send, replies to incoming guest messages in the inbox. Off by default. See AI replies to guests.
Fill in place coordinates — find latitude/longitude for places that are missing them.
Translate everything — the Translations page can fill every missing translation with AI.
Smart search — describe what a guest wants in plain language and let AI find matching listings.
Draft a request’s offer — from a request’s captured customer wishes, find matching listings, check availability and pricing, and create a draft multi-option offer with a suggested customer message. See Draft an offer with AI.

There’s also Ask Reynt, the in-console assistant — a full chat agent that can read your tenant’s data (bookings, listings, prices, customers, invoices, reports, and more), answer questions about the console, and propose changes for you to confirm. See The AI assistant.

Connect AI

AI features are configured under Settings → AI. A help panel on the page walks you through the same steps.

Provider — choose which vendor serves your account: Anthropic (USA) or Mistral (EU). This choice works with or without a key of your own — where the platform holds its own key for the vendor you pick, your requests run on that access and our data-processing agreement covers them. The page shows which providers are available that way, and where your requests are actually being processed right now.
API key (optional) — connect your own key to be billed directly by the provider instead. It also moves the data-protection duties to you — see Data protection below. If you pick a provider the platform holds no key for and add no key of your own, AI stays switched off for that provider: we will not quietly route your data to a different vendor instead.
Model — leave it on the default (Claude Sonnet 5 for Anthropic) for a good balance of quality and speed, or pick another model — Anthropic offers Opus 4.8, Opus 4.7, Sonnet 5, Sonnet 4.6, and Haiku 4.5 (faster, cheaper short texts); Mistral offers Mistral Medium, Large, Small, and Magistral Medium. You can also enter a custom model ID.
Custom instructions — optional guidance applied to generated content (tone, style, house rules). You can even generate a starting set from your own listings and configuration.
Smart search — a separate embedding provider (OpenAI, USA, or Mistral, EU), optional key, and model power the smart-search index. The same rule applies: the provider is selectable on the platform’s access, and a key of your own is optional. Switching the embedding provider or model re-indexes all your listings.

Once AI is available — on the platform’s access or your own key — the AI buttons light up across the console, most visibly on listing detail pages.

Cost and quotas

The key you use determines how spend is controlled:

Your own key (BYOK) — you pay your provider directly and there are no platform limits. It also changes who carries the data-protection duties — see Data protection below.
Platform key — when no key is connected, AI runs on the platform’s key subject to per-account limits (a cap on how many workflows run at once, per day, and per month).

The AI assistant has its own daily token allowance on the platform key (generous enough for normal use); connecting your own key lifts it entirely.

Each AI run records which key was used, so you can always see where cost landed.

Data protection

AI features send data to a language-model provider. You are the controller for your guests’ and owners’ data, so this is worth understanding before you turn them on.

What leaves the platform. Only what a given request needs, and only at the moment you trigger it. Depending on the feature that can include guest and owner master data (name, address), contact details (email, phone), booking data including special requests, and — only if you switch on AI replies to inbound guest messages — the text of those messages. Registration-form (Meldeschein) data is never sent to a model. Nothing is sent in the background: no call happens unless you or a feature you enabled makes one.

Who processes it. Whichever provider you selected under Settings → AI. Both are listed as approved subprocessors in Anhang 3 of the AVV:

Anthropic (USA) — the default. Each request is a transfer to a third country, on EU standard contractual clauses as the transfer basis.
Mistral (France, EU) — selectable on the platform’s access, no key of your own required. Because the vendor is established in the EU, your request is not a third-country transfer by us. Read the limit of that below.

Your content is not used to train models (AGB § 6(3)). The settings page shows which of the two is serving you, so you never have to infer it.

Tell your guests. Your own privacy notice has to mention that AI features process their data — and, if you are on the US provider, that this involves a transfer outside the EU. That duty is yours as controller, not ours (AVV § 5(3)).

If you bring your own key, the provider you choose is your processor, not ours. Our contracts and transfer safeguards do not cover that key. You then need your own Art. 28 agreement with that provider, your own transfer basis if they are outside the EU, and your own notice to data subjects (AVV § 6(6)). BYOK does not automatically make things safer — it moves these duties from us to you.

What choosing the EU provider does and does not settle. It removes the third-country transfer from us to the provider: Mistral is established in France, so that leg stays inside the EU, and it stays under our agreement with them rather than being pushed onto you. It does not amount to a promise of EU-only processing end to end. “EU company” and “EU-only processing” are not the same thing — a provider’s own sub-processors and hosting regions are documented per feature and change over time. Our contract therefore points at Mistral’s own transfer regime for those onward steps instead of promising residency we have not verified. If you need a hard residency guarantee rather than an EU-established processor, ask us before you rely on it.

One exception that no provider choice changes: searching the built-in help knowledge base. That index describes Reynt itself and is built centrally for all accounts, so it always uses the platform’s own embedding provider — your search text goes there regardless of the provider you picked. Anhang 3 lists this separately for that reason.

Reducing what goes out. Settings → AI has a pseudonymization option that replaces names, emails, phone numbers and street addresses with placeholders before anything reaches the model, and resolves them again before you see the answer. It is off by default: it costs the assistant some ability to match similar names and to inflect German salutations. It is a protective measure, not anonymization — the data still counts as personal data, and the provider remains a subprocessor. It is most worth enabling on a BYOK key, since that is the path our own safeguards do not reach — and worth considering on the US provider, where each call is a third-country transfer.

Monitor usage

The Settings → AI Usage page gives you visibility into consumption:

Usage tab — cards for workflows today, this month, and running now; totals for a chosen period (last 7 days, last 30 days, or this month) covering workflow count, tokens, and estimated cost; and a breakdown table by workflow type showing the key source, token counts, and cost.
Workflows tab — a history of every AI run with its type, status, related entity, and duration. Open any row to inspect the run’s steps in detail.