Reviews & reputation
Reviews from Google, Trustpilot and Tripadvisor, in one shape and one bill
Three review platforms mean three integrations, three shapes and three review-of-the-terms conversations with legal. The data is the easy part; the part that stalls is proving where it came from.
Reputation data is personal data, and procurement knows it
A review is written by an identifiable person about an identifiable business. Pulling a thousand of them into your warehouse is a processing activity with a lawful basis, a retention period and a sub-processor chain — which is why review pipelines get blocked by legal far more often than by engineering.
It is also the category where scraped supply is most common, and the one where "where did this come from?" has the least satisfying answer. If the honest answer is a proxy pool, the risk is not the vendor’s. It is in your product.
- Three sources, three field names for the same rating, three pagination models.
- Sentiment computed by three different tools is not comparable across sources.
- A summary without citations is an opinion; with them it is evidence you can show a franchisee.
One envelope across the sources
Step 1
The same request shape everywhere
Google, Trustpilot and Tripadvisor reviews arrive through routes with the same envelope, the same error contract, the same credit headers and the same idempotency behaviour. One integration, three sources.
Step 2
Find the place, then read it
Search endpoints resolve a business on Trustpilot or Tripadvisor, and business info resolves a Google profile, so you are not asked to bring platform-specific identifiers you do not have.
Step 3
Summarize with the reviews in hand
POST /v1/ai/review-summary takes reviews you already fetched and returns themes and complaints that point back at the reviews they came from — synthesis on models we host in the EU, so the review text does not travel to a model vendor.
The call, in full
curl -X POST https://api.dataswap.io/v1/reviews/google \
-H "Authorization: Bearer $DATASWAP_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "keyword": "IKEA Alfragide" }'Google reviews for one business profile. This route answers 200 or 202 — see the limits below.
The calls, and what they cost
| Endpoint | What it does | Price |
|---|---|---|
POST /v1/reviews/google | Reviews from a Google business profile. | 2 credits($0.0040) |
POST /v1/reviews/trustpilot | Reviews from a Trustpilot profile. | 2 credits($0.0040) |
POST /v1/reviews/tripadvisor | Reviews from a Tripadvisor listing. | 2 credits($0.0040) |
POST /v1/business/info | The Google business profile itself: hours, categories, rating and attributes. | 2 credits($0.0040) |
POST /v1/amazon/reviews | Reviews on an Amazon listing, for products rather than places. | 2 credits($0.0040) |
POST /v1/ai/review-summary | Themes, complaints and praise across reviews you pass in, with citations. | 3 credits |
Reads are priced per call at the depth you ask for, so a shallow sample across many locations costs less than a deep read of one. The summary is a separate route because it is a separate decision: fetch every day, summarize when someone is going to read it.
1 credit = $0.002 on every plan. Full plans and prepaid packs are on the pricing section.
Where this stops
Review reads answer 200 or 202. When the upstream read exceeds the handoff window you get a job_id and collect the same body from GET /v1/jobs/{id}; no credits are charged until there is a result, and resubmitting starts a second job with a second reservation.
It reads reviews; it does not answer them. There is no posting, replying or moderation — this is a data API, not a reputation console.
Software review platforms such as G2 and Capterra are not covered. If that is your category, this page is the wrong shelf.
Coverage of a given place depends on the source having a profile for it. A business with no Trustpilot presence has no Trustpilot reviews to read, and the response says so rather than inventing an empty rating.
We do not deduplicate the same human across platforms, and we do not attempt identity resolution on reviewers. Doing so would create personal data that did not exist before the call.
Questions we get asked
- Is scraping involved?
- No. The data comes from licensed upstream providers, and we do not build anti-bot evasion, CAPTCHA solving or proxy rotation. For review data specifically that is the difference between a pipeline your legal team can sign off and one it cannot.
- Are reviews personal data, and do you offer a DPA?
- Reviews generally are personal data under the GDPR, and yes: a signed data processing agreement is available from the Starter plan upward, with sub-processor categories and transfer terms published in advance.
- Do you store the review text you return to me?
- No. We do not store the responses we serve. We keep a usage record of the call for billing, and for 30 days the request arguments and originating IP so you can inspect the call in your dashboard; after that both are erased.
- Does the summary send review text to OpenAI or Anthropic?
- No. Synthesis runs on models we host ourselves inside the EU, so no model vendor is added to your sub-processor chain and the receipt records inference: eu-only.
- Can I get sentiment scores as well?
- Content analysis routes for sentiment and rating distribution exist in the reference and are billed separately. We keep them apart from the review read on purpose: fetching and scoring have different cadences and different prices.
Keep reading
Put the prism to work.
Create an account and get your API key — no card required. Start on the six live bands and the twelve inference tools; the eight new capabilities land on the same key as they ship.