Guide — website-type classification

Blog vs OTA vs operator: how to tell a real publisher apart from the rest

A partnerships list full of OTAs, aggregators, operators, and tourism boards wastes your team's best outreach hours. Website-type classification separates genuine content publishers from everything that merely looks travel-relevant — before a single email goes out. This guide shows the taxonomy, the signals that split each class, and the anonymized run that proves it holds at scale.

0
travel-related domains isolated in one run
0
precision on a reviewed sample
5
website types the classifier keeps apart
Why classification decides the conversation

Keyword-relevant is not the same as recruitable

A keyword filter treats every travel-flavored domain as a lead. Half of them cannot enter your affiliate program at all — an online travel agency is a competitor, a tour operator sells its own inventory, a tourism board never monetizes. Sorting by website type is the difference between an outreach list and an inbox full of polite declines.

Fewer dead-end emails

Operators and OTAs bounce your partner pitch by definition. Removing them up front means every send lands on a site that can actually carry your links.

Right opening line

You address a publisher as a publisher — commission, content fit, audience — not with a pitch that reads as if you mistook them for a booking engine.

The uncontested long tail

Once the aggregators are filtered out, what remains is the independent editorial layer your rivals never bother to reach — where recruiting capacity is still open.

The taxonomy

Five website types, only one of them is your publisher

Every travel-relevant domain sorts into one of these classes. The classifier reads what the site actually does — publishes content, sells inventory, aggregates other sellers, promotes a destination — not what its meta tags claim. Only the first class is a recruitable affiliate.

Recruit

Content publisher

Independent blogs, review hubs, guide sites, and destination newsletters that publish editorial and monetize through affiliate links or ads.

Tells: article structure, author bylines, evergreen guides, affiliate or ad monetization, no own booking flow.
Exclude

OTA

Online travel agencies that sell flights, hotels, or packages directly through their own checkout. Structurally a competitor, not a partner.

Tells: booking widget, date-picker search, cart and payment flow, inventory pricing, "book now" as the primary action.
Exclude

Aggregator / metasearch

Comparison engines that pull listings from many sellers. They may look editorial but run on a feed, not on writers.

Tells: live price tables, deep-link redirects to sellers, thin surrounding copy, filters over articles.
Exclude

Operator / supplier

Tour operators, hotels, airlines, and activity providers selling their own product. They are the advertiser's peers, not its affiliates.

Tells: own inventory, availability calendar, brand-first navigation, "our tours / our rooms" language.
Exclude

DMO / tourism board

Official destination marketing organizations. Authoritative and content-rich, but non-commercial and unable to join an affiliate program.

Tells: government or .gov-style ownership, no monetization, "official visitor guide" framing.
Exclude

Thin / MFA site

Made-for-advertising and spun-content domains with no genuine audience. They pollute a list and damage sender reputation.

Tells: templated pages, no author, stale or generative copy, ad density far above editorial norms.
The distinctions that trip people up

Two pairs that a keyword filter always confuses

Most misclassification comes from two look-alike pairs. Naming the discriminating signal for each is what turns a fuzzy guess into a reliable call.

Publisher vs aggregator

Both review options and link out. The split is the engine: a publisher writes opinionated editorial and places affiliate links inside it; an aggregator renders a live feed of sellers with minimal original writing. Look for authorship and article depth, not for the presence of outbound links.

Publisher vs operator

A publisher recommends things it does not sell; an operator sells its own product. The decisive signal is own-commerce — a checkout, availability calendar, or inventory pricing. If the site takes the booking itself, it is a supplier, not a partner.

The instrument

The signals that separate the classes

Each domain is read against a fixed vocabulary of website-type signals, every one extracted from what the page actually contains and shipped with a confidence score. Website type is derived from the combination — not from a single keyword.

Article / editorial structure Named author or byline Evergreen destination guides Affiliate links present Display ad monetization Newsletter / subscriber base Content freshness Individual or small-team run Own booking / checkout (OTA/operator) Availability calendar (supplier) Live price feed (aggregator) Non-commercial / official body (DMO) Thin or spun content (MFA)

Need a distinction that isn't here — "runs its own OTA on a subdomain", "affiliate blog attached to a supplier brand"? Custom website-type rules are defined per engagement and calibrated the same way.

The method

How the classification actually runs

Website type is not guessed from a domain name. It is read from the site itself, scored, and reconciled against the exclusion rules before anything reaches your list.

1

Pull the vertical

Every travel-associated domain in the 120M+ classified-domain database enters the pool — the whole class, not a directory's members.

2

Read each site

An LLM reads the live page and extracts the website-type signals — booking flow, feed, byline, monetization — each with its own confidence score.

3

Assign the type

Signals resolve to one of the five classes. Own-commerce and live-feed signals force an exclusion regardless of how editorial the wrapper looks.

4

Calibrate, then keep

A reviewed sample confirms the split holds before the full run, and only the content-publisher class lands on your recruitment list.

Why order matters: a single "book now" widget outweighs a hundred travel keywords. Deriving type from structural signals — not vocabulary — is what keeps competitors and suppliers out of the final CSV.
Proven at scale

The classification, tested on a whole vertical

In a large-scale production run on our own classification platform, the pipeline read the travel vertical end to end, sorted every domain by website type, and kept only content publishers. The client reviewed a sample of the output against their own judgment.

0
domains in the source dataset
0
travel-related domains identified
0
domains ICP-screened in one run
0
precision on a reviewed sample
30M
source domains
1.73M
travel-related
1.2M
ICP-screened
~96%
precision confirmed

Numbers from one anonymized engagement. Multi-country travel programs sit on millions of candidate domains; a single-vertical run typically yields tens of thousands to hundreds of thousands of qualifying publishers once suppliers and aggregators are removed (estimate).

Side by side

What the reader sees vs. what the classifier reads

The surface impression of a travel site is a poor guide to whether it can partner with you. The classifier ignores the impression and reads the structure.

Website typeCan it carry your links?Looks travel-relevant?Decisive signal
Content publisherYes — recruitYesEditorial + affiliate/ad monetization, no own checkout
OTANo — competitorYesOwn booking and payment flow
AggregatorNo — feed, not writersYesLive price feed, deep-link redirects
Operator / supplierNo — supplierYesOwn inventory + availability calendar
DMO / tourism boardNo — non-commercialYesOfficial body, no monetization
Questions

Website-type classification — FAQ

Isn't a domain's category enough to tell publishers from OTAs?
No. "Travel" is a topic, not a business model. An OTA, a tour operator, and an independent blog all classify as travel by topic. Website-type classification adds the second axis — what the site does with that topic — which is the axis that decides whether it can join your program.
How do you handle a hybrid — a blog with a small booking widget?
Hybrids are exactly what calibration is for. By default, any own-commerce flow tips a site into the supplier class. If you want to keep affiliate blogs that also sell a minor product, we adjust that rule during the reviewed-sample phase and confirm the new split before the full run.
What does the ~96% precision figure actually describe?
In one anonymized travel engagement, the client reviewed a sample of the delivered publishers and judged roughly 96% to be correct content-publisher fits. It is a measured figure from a single run, not a guarantee. Your own precision is measured on your reviewed sample before we commit to the full run.
Does this work outside travel?
Yes. Every vertical has its own look-alikes — retailers vs. review blogs, brokers vs. finance publishers, vendors vs. software reviewers. The same website-type axis separates the recruitable publisher from the competitor or supplier in each of the 30 industries we cover.
How do I see the classification on my own vertical?
Request a free pilot. We configure the full pipeline for your ICP, sort your vertical by website type, and deliver the first 20 qualified content publishers — plus a projected full-run yield. You see real names and the type split before committing to anything.

See your vertical sorted by website type

The free pilot runs the full pipeline on your industry, separates publishers from OTAs, aggregators, operators, and DMOs, and delivers your first 20 qualified publishers plus a full-run projection. No cost, no obligation.

Request a Free Pilot