Lookalike websites: finding more sites like the ones a client already trusts
The short answer
Lookalike websites are publishers whose content closely resembles a few seed sites a client already trusts. Buyers find them by comparing what the seed sites' articles mean, not just their category label, searching each theme separately when the seeds differ, and then checking every match for quality, suitability, selling path and audience before it joins a plan.
Key takeaways
- A lookalike website matches the content of a seed site; a lookalike audience matches people, and the two answer different questions.
- A category label alone groups sites that share a label but not a subject; comparing the meaning of sampled articles finds closer matches, but only the matching articles make those matches checkable.
- Seeds on different themes must be searched one theme at a time, or the results drift toward sites that match none of them well.
- Similar content does not mean similar quality: every lookalike still needs MFA, suitability, seller and audience checks.
- Large generalist sites and networks of near identical domains are two ways a lookalike list goes wrong.
What are lookalike websites in advertising?
Definition
- Lookalike websites
- Publishers whose articles cover the same subjects, in a similar depth and voice, as one or more seed sites that a client or planner already knows and trusts. They extend a short list of proven sites into a plan with enough scale to deliver.
The request can arrive in a single sentence: "We love running on these three sites, find us more like them." The client is describing a feeling about a kind of publication: the subjects it covers, how seriously it treats them and who reads it. The planner's job is to translate that feeling into evidence that can be checked.
Lookalike websites are not lookalike audiences. Google describes its Lookalike segments (opens in a new tab) as groups of people who share characteristics with an existing seed list built from the advertiser's first-party data, documented for Demand Gen campaigns. Google Ads stopped generating the older similar audiences (opens in a new tab) on May 1, 2023. A lookalike website list uses no data about people at all. It selects places by their content.
How do category, keyword and meaning matching compare for finding lookalike sites?
Content categories are a shared vocabulary for describing pages. The IAB Tech Lab Content Taxonomy (opens in a new tab), listed at version 3.1 with a December 2024 release date, names contextual targeting and brand safety as its typical uses. On its own, a top level label can be too coarse for lookalikes. A board game review magazine, a toy store's blog and a casino games site can all carry a games label.
Meaning based matching compares what the articles actually say. Sentence embedding methods such as Sentence-BERT (opens in a new tab) turn text into vectors whose closeness reflects closeness of meaning, so a review of family strategy games sits near an article on picking a first board game for children even when they share few words. Multilingual versions (opens in a new tab) map a sentence and its translation to nearly the same point, which lets an English seed find a French Canadian lookalike.
| Approach | What it compares | Strength | Weakness |
|---|---|---|---|
| Category match | The seed's top level content label | Fast, uses a shared industry vocabulary and easy to explain to a client | Returns large, loosely related sites that share the label but not the subject |
| Keyword overlap | Words that appear on the seed's pages | Transparent; good when the seed's subject has distinctive names or terms | Finds sites that use the same terms in another sense and misses other languages |
| Meaning match on sampled articles | What several of the seed's articles are about, compared with several of each candidate's | Finds specialist and non English publications on the same subjects | Depends on the model and the sample; can favor well written pages on a nearby subject; needs the matching articles shown to be checked |
How should a planner choose the seed sites?
The seeds define the search, so weak seeds produce weak lookalikes. This guide works with one to three seeds. More than that can blur what the client actually values, so ask the client which two matter most.
What makes a good seed site
- The seed is a real publisher with its own editorial articles, not a store, brand site or forum index.
- Most of the seed's recent articles are on the subject the client cares about, not just the one story they remember.
- The seed passed the same quality checks the lookalikes will face, including a made-for-advertising review.
- The planner knows why the client likes the seed: its subject, its tone, its local audience or its depth.
- The seeds either share one theme or are labeled with the separate themes they represent.
- A specific section or article URL is used when only part of a large site fits the brief.
That last point matters for general sites. If the client's favorite is the games column of a national newspaper, seed with that column's articles. Seeding with the whole newspaper finds other newspapers, which may not be what the client meant.
Why should seeds on different themes be searched one theme at a time?
Clients can hand over seeds that look alike to them but cover different subjects: a personal finance blog, a home renovation magazine and a parenting site, all read by young homeowners. If a search blends all three into one target, the closest matches are sites that sit halfway between them, such as general lifestyle portals that cover everything thinly. None of the client's three themes is served well.
What is a reliable method for building a lookalike website list?
Seven steps from seed sites to an approved lookalike list
- Confirm the seeds and the reason. Write down each seed, the theme it stands for and what the client values in it.
- Sample the seeds. Read several recent articles from each seed and note what they are about in a sentence each.
- Search each theme separately. Look for sites whose sampled articles mean the same thing, in every language the brief allows.
- Judge whole sites. Keep candidates where most sampled articles match, and drop broad sites where only a sliver of the sample fits.
- Run the standard checks. Review page quality, made-for-advertising risk, suitability, the ads.txt (opens in a new tab) selling path and audience in the target country.
- Collapse networks. Treat sites that share one owner's ad or analytics codes as a single entry.
- Label and approve. Record which seed each site resembles and the articles that prove it, then send the list for approval.
How does PlacementGraph find sites like the ones a client likes?
Which pitfalls weaken a lookalike website list?
| Pitfall | How it shows up | Fix |
|---|---|---|
| Generalist drift | Broad portals rank high because a handful of their articles match | Score the share of sampled articles that match, not the best single article |
| Network clones | Several near identical domains with the same layout fill the list | Group sites by shared ad and analytics codes and keep one per network |
| Copycat quality | Templated or rewritten sites match the seed's subject closely | Run the made-for-advertising review on every lookalike, not only on the seeds |
| Popularity bias | Only large, well known sites survive because small ones lack traffic data | Show reach as information and keep relevance as the ranking signal |
| Language mismatch | A site matches in meaning but publishes mostly in a language the brief excludes | Detect the language of each sampled article, not the domain ending |
| Seed overreach | A whole large site was used as the seed when only one section fits | Seed with the specific section or articles the client means |
Popularity bias deserves extra care. Public datasets such as the Chrome UX Report (opens in a new tab) only include origins that pass an undisclosed popularity threshold, so a small specialist publisher can be missing from the data while still having loyal readers. Missing reach data is a gap to note, not a reason to drop a strong lookalike.
What does a lookalike search look like on a real brief?
Fictional example
A fictional board game publisher in the United States
Meeplewright Games, a fictional publisher of family board games, gives its agency three favorite sites: a tabletop game review magazine, a family game night blog and a guide to game cafes in the Midwest. The planner notices two themes, game reviews and family activities, and runs one lookalike search for each.
The review search returns specialist tabletop publications, including a French language Quebec site on game design that the client had never seen. The family search returns parenting and weekend activity sites with regular game features. A national entertainment portal with two board game stories drops to the bottom because most of its sampled articles cover film and television.
Four domains with the same template and shared ad codes collapse into one entry, then fail the page quality review. The final list labels every site as game reviews or family activities, with two matching articles linked beside each.
Guides in this hub
Every guide in the Lookalike websites hub, each written to answer one question in full.
- FrameworkAdjacent audiences advertising: a framework for testing topics next to the coreA framework for testing adjacent topics next to a core site list: how to pick them, size a small budget, keep them separate and decide when one joins the plan.
- GuideHow do B2B advertisers find and judge trade publications?How B2B tech advertisers find and judge trade publications without account data: which sites reach buyers, how to read gated sites, and how to size the list.
- GuideHow should advertisers judge beauty websites before buying ad space?How to judge beauty and fashion sites for ads: sponsored post disclosure, influencer led blogs, cosmetic and drug claims, and growing a list from trusted seeds.
- How-toHow do you audit a site list a client hands you?How to audit a list of favorite sites a client hands you: the checks every site must pass, what to do when one fails, and how to explain removals clearly.
- GuideHow should advertisers judge DIY and craft websites before buying?How to judge DIY and craft sites for ads: original tutorials against copies, image search and pin traffic, ad heavy layouts, craft seasons and lookalike search.
- GuideHow should advertisers vet entertainment websites before buying?How to vet film, TV, music and celebrity sites for ads: editorial signals, piracy and lyrics rights checks, gossip tolerance, family fit and lookalike search.
- How-toHow do you expand a programmatic site list that underdelivers?A site list that underdelivers can grow without diluting fit. Learn which criteria to widen, which checks never move, how to add a market and report the change.
- How-toHow do you find local websites for advertising in each city?A city by city method for finding local websites for advertising from one good local seed, in English and French, with checks that prove each is a publisher.
- How-toHow do you find publishers in your own referral traffic?Use your own analytics referral rows to find publishers your visitors already read, keep personal data out, filter the noise and turn the best sites into seeds.
- How-toHow do you find similar websites, starting from one seed site, and check them?A step by step method to find similar websites from one seed site: describe its articles, gather candidates, sample their pages and check each one.
- How-toHow to find similar YouTube channels from one the client already likesExpand from one YouTube channel a client likes to similar channels using topics, video meaning and audience signals, and screen out re-uploads before the plan.
- How-toHow do you find websites by topic when there is no seed site to start from?A topic first method for building a site list from scratch: map the topic, search for specialists, sample articles, set aside stores and log every find.
- GuideWhich food publishers are worth advertising on, and how do you tell them apart?How buyers judge food and recipe publishers: signs of original testing, why recipe pages run heavy ad loads, how to spot recipe farms and find trusted sites.
- GuideHealth websites for advertising: how to tell a trustworthy publisher from a content farmWhat makes a health website trustworthy for advertisers: medical review, sources, dates and corrections, plus how to spot content farms and find similar sites.
- GuideHome improvement websites advertising: telling publishers from stores and directoriesHow to judge home improvement websites for advertising: separate publishers from stores and contractor directories, test renovation advice, plan for seasons.
- GlossaryWhat are lookalike audiences, and how do site lookalikes differ?What lookalike audiences are, how a seed list becomes a modeled group of people, how site lookalikes differ, and the personal data each approach relies on.
- ExplainerWhat is lookalike drift, and when should expansion stop?What lookalike drift is, why repeated expansion rounds wander off theme, the signs a planner can measure, a stop rule and how to reset a drifting site search.
- How-toWhen should you rerun a lookalike search, and how do you show what changed?When to rerun a lookalike website search, which seed changes force an early refresh, and how to show added, kept and removed sites to the client.
- ChecklistWhat should a lookalike site list pass before approval?A final checklist for approving a lookalike site list: theme fit, seed tracing, quality, sellers, market audience, duplicates, review flags and who signs off.
- How-toHow do you find Canadian lookalikes of US seed sites?How to find Canadian lookalike websites from US seed sites, find French counterparts of English seeds, and confirm each match has a real Canadian audience.
- GuideHow should a buyer judge magazine websites for advertising?How buyers judge magazine websites for advertising: who runs each section, how web formats differ from print, which titles suit niche brands, how to find more.
- FrameworkHow do negative seed sites steer a lookalike search away?A framework for negative seed sites: when to give a lookalike search examples of what the client does not want, how many to use and how to record them.
- GuideHow should buyers judge new websites for advertising fairly?How buyers judge new websites for advertising when history is thin: the checks that work on day one, what to watch for, and how to run a small, fair test.
- GuideHow do you find the niche publishers that broad site lists leave out?How to find niche publishers that broad site lists miss: why they go unseen, the vertical signals of a specialist outlet, where to look and how to check them.
- GuideParenting websites advertising: how to judge a site written for parentsHow to judge parenting websites for advertising: who a site is written for, which topics need suitability review, and how to find more trusted parenting sites.
- FrameworkHow do you turn a customer persona into a site list?A framework for turning a customer persona into content topics, formats and publisher types, with the persona claims to drop and a test for each site.
- GuideHow should advertisers judge personal finance websites before buying ad space?How to judge personal finance sites for ads: affiliate card and loan offers, editorial independence statements, scam adjacency, and growing a trusted list.
- GuideHow should advertisers judge pet websites before buying ad space?How to judge pet sites for ads: checking veterinary reviewers, recall coverage, breed forums and pet topics to review, then growing a list from trusted seeds.
- GuideHow should advertisers judge product review sites before buying ad space?How to judge product review sites for ads: real testing, affiliate disclosure under the FTC Endorsement Guides, review farms and reviews of your own brand.
- GuidePublisher discovery: where buyers find publishers they have never usedPublisher discovery sources for media buyers: associations, award lists, citations, link graphs, newsletters and local directories, and how to log each one.
- GuideHow should advertisers judge science websites before buying?How to judge science news and explainer sites for ads: research sourcing checks, preprints and retractions, why university sites are not inventory, and fit.
- How-toHow do you find similar websites from one example article?Seed a lookalike website search with one to three example articles: when article seeds beat site seeds, how to pick them and how to check every result.
- GuideHow do you choose seed sites that lead a lookalike search to the right places?How to choose seed sites for a lookalike search: where seeds come from, the failure modes that mislead it, a scoring table and questions for the client.
- ExplainerGoogle's similar audiences ended in 2023. What replaced them?What changed when Google stopped generating similar audiences in 2023, what replaced them in Google Ads, and the site level options that use no data on people.
- GuideSports websites advertising: judging sports publishers and keeping betting apartHow to judge sports websites for advertising: team blogs, league news and fantasy sites, keeping sports coverage apart from betting, and finding similar sites.
- GuideHow should advertisers judge technology news sites before buying ad space?How to judge technology news sites for ads: spotting press release rewrites and AI written posts, consumer and enterprise tech, and topics that need review.
- GlossaryWebsite categorization: how sites get labels, and when a domain label misleadsHow websites get categories, why a domain label can mislead buyers on mixed sites, how many labels one site carries and how labels feed lookalike searches.
- GlossaryWhat does website similarity mean, and what does each measure prove?Two sites can be similar by topic, meaning, audience, links or ownership. Learn what each measure proves, what it misses and which one needs personal data.
- GuideHow should advertisers judge wedding websites before buying ad space?How to judge wedding sites for ads: planning guides and real wedding features against vendor directories, disclosure checks, seasonality and lookalike search.
- How-toWho owns a website, and how can an ad buyer find out?How ad buyers find who owns or runs a website: about pages, ads.txt OWNERDOMAIN, sellers.json, US and Canadian business registries, and what to do with gaps.
Frequently asked questions
How many seed sites do I need to find lookalike websites?
One to three can be enough, as long as each seed is a strong example of what the client values. A single strong seed gives a clearer target than five loosely related ones. When the client offers many, ask which two matter most and why, then use that reason to guide the review.
Can lookalike websites be found in other languages?
Yes, when the comparison works on meaning rather than words. Multilingual embedding models place a sentence and its translation close together, so an English seed can surface French or Spanish sites on the same subject. Check each candidate's language from its articles and keep only the languages the brief allows.
Do lookalike websites need the same checks as any other site?
Yes. Similar subjects say nothing about ad density, paid traffic, suitability or who sells the inventory. A site can copy the subjects of a good publisher without its standards, so every lookalike should pass the same quality, suitability, seller and audience review as the rest of the plan.
What should I do with sites that are close but not quite on theme?
Keep them apart from the core list as adjacent ideas. A puzzle and crossword site found from a board game seed may suit a gift guide angle. Show these separately, explain the connection, and let the client decide whether to test them with a small share of budget.
Sources
- Use Lookalike segments in Demand Gen campaigns as a Targeting Constraint (opens in a new tab)Google Ads Help
- Changes to audience targeting: Google Ads will no longer support similar audiences (opens in a new tab)Search Ads 360 Help
- IAB Tech Lab Content Taxonomy (opens in a new tab)IAB Tech Lab
- Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks (opens in a new tab)Nils Reimers and Iryna Gurevych, EMNLP 2019
- Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation (opens in a new tab)Nils Reimers and Iryna Gurevych, EMNLP 2020
- CrUX methodology (opens in a new tab)Google Chrome for Developers
- ads.txt (Authorized Digital Sellers) (opens in a new tab)IAB Tech Lab
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.