Lenso.ai is an AI reverse image search engine with five separately trained categories — faces, places, duplicates, related and similar — plus alerts, advanced filters and a DMCA takedown helper. Face search is run by a partner and is unavailable in the EEA, the UK, Argentina and Illinois.
Used this tool? Rate it
Used this tool? Rate it
Lenso.ai is an AI reverse image search engine that lets you start from a picture rather than a keyword, searching for places, people and duplicates across what it describes as billions of images from around the web. You drop, paste or upload a photo, and the engine returns pages where visually matching images appear.
That basic description is true of several tools. What distinguishes this one is structure: the product is organised into five search categories — faces, places, duplicates, related and similar — each driven by a separately trained model, so that every category returns results that differ in kind rather than merely in ranking. Searching a photograph of a building under "places" and under "similar" gives you two genuinely different answers: one tries to identify the location, the other looks for visually comparable images.
Two facts shape everything else in this review, and neither is prominent on the marketing pages.
The first is corporate. The operator is a Polish joint-stock company, LENSO AI SPÓŁKA AKCYJNA, registered in Wrocław under KRS number 0001106624 with fully paid share capital of 1,040,000 PLN. That places the company squarely inside the EU regulatory perimeter, which matters more here than for most tools, because the second fact concerns who actually runs the face search.
And that second fact is the single most important thing the marketing pages do not tell you: Lenso does not run the face search itself — the privacy policy states plainly that the controller performs no facial search and that it is performed entirely by EYEMATCH AI PTE. LTD, a Singapore company to which your photo is forwarded. Independent confirmation exists: the trade publication Biometric Update reported on 1 September 2024, under the byline of Abhishek Jadhav, that Lenso.ai had formed a strategic partnership with EyeMatch.ai to build a consumer-facing facial recognition search engine. So when you use the People category, you are entering into a relationship with two companies, in two jurisdictions, not one.
The category split is the product's real architecture, and understanding it is the difference between useful results and frustration. "Duplicates" answers "where else does this exact image appear". "Similar" answers "what looks like this". "Related" broadens further into thematic neighbours. "Places" attempts geographic identification. "People" attempts to match a face.
Pick the wrong category and you will conclude the engine is weak when in fact you asked the wrong question. If you are chasing an image thief, "duplicates" is the category; "similar" will drown you in visually adjacent but irrelevant photographs.
The face search is pitched at finding where photos of your own face appear online, with the company claiming high precision through accurate matching of facial details — a claim that carries an asterisk on the homepage itself. That asterisk matters, because the free-trial framing around face search is conditional rather than open-ended.
The second advertised use is copyright and provenance: the company says the tool excels at finding exact duplicates and tracing original sources, while acknowledging that what it displays are photos hosted elsewhere. That last clause is an honest one and worth keeping in mind — Lenso is an index pointing outward, not a repository. Read carefully, it also sets the ceiling on what image copyright protection can mean here: the tool locates reuse, it does not police it.
For paying users there is a depth setting. Research Mode widens the People and Duplicates categories to 10,000 results, which the company describes as ten times the normal ceiling, at the cost of a few extra seconds and with access to results the standard search does not surface. The phrasing there is significant: the standard search is not showing you everything it found. If your first search came back thin, the ceiling rather than the index may have been the constraint.
Depth without sorting is noise, which is what the filters are for. The advanced filters cover five axes: domain language with include and exclude lists, the date range in which Lenso indexed the photo, image size, specific domains to include or exclude, and a keyword that must appear in the page title.
The date-range filter deserves a note of precision: it filters by when Lenso indexed the image, not when the image was published. For provenance work that distinction is material — index date is a lower bound on discovery, not evidence of origin.
Finding a stolen photo is only half the job. DMCA Assist walks you through a DMCA takedown request, gathering the image links and sources you select, dropping them into an email template and generating a ready-to-copy message addressed to whoever hosts the photo. It is a drafting aid, not a legal service, and it does not send anything for you.
Alerts turn a one-off search into monitoring, emailing you when new matches appear, with ceilings of 10, 50 and 500 alerts across the three tiers. This is the feature that converts the product from a lookup tool into an ongoing service — and, as covered below, the one with the most significant privacy consequence.
The headline personal use case, and the reason many people arrive here. Worth checking the regional restrictions before you invest time, because a large share of readers of an English-language page will be in a region where this specific category is switched off.
The duplicates category plus DMCA Assist is a coherent workflow: find duplicate images, select the ones you did not license, then generate the takedown text. For working photographers, this pairing is the strongest argument for the paid tiers. For working photographers, this pairing is the strongest argument for the paid tiers.
Journalists, moderators and anyone assessing a suspicious photo can use duplicates to find earlier appearances of the same image. An image that surfaces in an old, unrelated context is a strong signal that a current claim about it is false.
The combination of large result ceilings, domain filters and index-date ranges is aimed squarely at investigative work. Biometric Update's framing of the category, discussed below, is essential reading before you use it in this mode.
The places category serves travel research, real-estate verification and the ordinary curiosity of "where was this taken".
The API tier serves teams building image lookup into their own applications — though, as covered under pricing, the entry cost for that tier is far higher than the consumer tiers would lead you to expect.
The free tier works with or without an account but is restricted to personal, non-commercial use, returns a portion of its results in a blocked state, and may show you ads. This is enough to judge whether the engine finds anything relevant at all — the blocked results still tell you how many matches exist, which is the number that determines whether paying is worthwhile.
Before searching, decide which of the five questions you are actually asking. This single step does more for result quality than any paid upgrade.
The paid currency here is "unlocks" — revealing the source link behind a result. Since quotas are 50, 500 or 5,000, and each unlock is spent permanently, scan the thumbnails first and unlock selectively.
If your results are numerous but irrelevant, reach for domain and language filters before Research Mode. Depth multiplies both signal and noise; filters remove noise only.
An alert is the right tool for ongoing monitoring — and, as the privacy section explains, the one action that changes how long your image is kept. Set it for images you genuinely want watched.
A Chrome extension lets you right-click any image on any site and search it directly, and the privacy policy states that it only reads page content when you actively invoke that feature and does not track you across other sites.
Test with an image whose answer you already know. Search a photo you have personally published somewhere. If the engine does not find that known copy, its index does not cover your corner of the web, and no subscription changes that.
Check regional availability before paying anything. If you are in the EEA, the UK, Argentina or Illinois, the face search will not work for you. Confirm this first if People is the category you came for.
Understand that unlocks are the real meter. Comparing tiers by price alone is misleading; compare by unlocks, because that is the quota that runs out.
Do not upgrade casually. As detailed under pricing, upgrades are charged in full immediately and are non-refundable even when clicked by accident.
Treat index dates as discovery dates. When building any provenance argument, never present a Lenso index date as a publication date.
Use the opt-out route if you are the subject rather than the searcher. The same product that finds your face also provides the mechanism to remove it, and that mechanism is free and fast.
Photographers, illustrators and stock contributors. The clearest fit. Duplicates plus alerts plus DMCA Assist maps directly onto the licensing-enforcement workflow.
People auditing their own online image footprint. A strong fit where the face category is available, and it comes with a working removal path.
Journalists, fact-checkers and moderators. Duplicate tracing is a core verification technique and this tool does it well.
Investigators and OSINT practitioners. The depth and filtering are built for this, with the ethical caveats in the limitations section.
Developers embedding image search. Served, but only at a tier priced far above the consumer plans.
Who should look elsewhere: anyone in a restricted region who specifically wants face search; anyone expecting removal from Lenso's index to remove an image from the internet; and anyone who wants a free, unlimited tool, since the free tier deliberately blocks a share of results.
Lenso.ai is a web product, used primarily in the browser at lenso.ai, with a Chrome extension as the second entry point. There are no native mobile or desktop applications, and the company publishes no app-store presence.
Machine access is a separate track. The API comes in two flavours — a Face Search API corresponding to the People category, and a Reverse Image Search API covering Duplicates, Places, Related and Similar.
Service levels are published rather than implied, which is unusual for a consumer tool and welcome. The published SLA promises 95 percent monthly availability on the free tier and 98 percent on paid tiers, and — importantly — explicitly excludes the People search engine because it is run by another provider. Read that exclusion carefully: the most sensitive feature in the product is also the one carrying no availability commitment.
At the time of writing, the three tiers were priced at $19.99 per month for Starter, $69.99 per month for Professional and $2,800.00 per month for Developer, each marked plus tax, with a yearly toggle advertising 20 percent off.
That Developer figure is not a typo, and it reframes the product. Forty times the Professional price places the API firmly in institutional territory rather than indie-developer territory.
What separates the tiers is quota rather than search quality: 50, 500 and 5,000 unlocks respectively, alongside 10, 50 and 500 alerts, with Research Mode and DMCA Assist arriving at Professional and API access only at Developer. Note what is not in that list — the underlying index and matching quality are the same at every tier.
Payments run through Paddle.com Inc as the external payment intermediary, which is why invoices and receipts arrive from Paddle rather than from Lenso.
The refund terms are strict and worth reading before subscribing. There is a narrow right of rescission: within 14 days of payment, if you have not used a single unlock and have no active alert, you can cancel the agreement from the user panel — but using one unlock or setting one alert forfeits it. In practice this means your evaluation must happen before you touch the product's core function.
Two further terms deserve attention. Upgrading charges you the full amount of the new plan immediately with no credit for the old one, and the terms spell out that this applies even when the upgrade was unintentional, including a mistaken click on the upgrade button. And cancelling does not refund unused balance: funds covering time beyond the current billing period are not returned, and the agreement simply ends on the last day of that period.
Google Lens is free, enormous in coverage and the default first stop, but deliberately does not offer face matching and gives you no takedown workflow.
TinEye is the veteran of exact-duplicate search, with a well-regarded index for finding copies and modifications, and no face search at all.
PimEyes is the direct comparison for face search specifically, and the tool most often named in press coverage of this category — including in the Biometric Update piece cited below.
Yandex Images is widely regarded as strong at visual and face-adjacent matching and is free, with obvious jurisdictional considerations of its own.
For most copyright work, the honest recommendation is to run a free tool first and reach for a paid tier only when the free ones come back empty.
The People category is geographically restricted and is not offered where such services are not legally permitted, with the policy naming EEA countries, the UK, Argentina and the US state of Illinois. This is the most consequential limitation in the product and it is disclosed in the privacy policy rather than on the pricing page. If you are in Europe, the headline feature is simply off.
Restating the structural point, because it affects your risk assessment rather than just your curiosity: Lenso forwards your photo to a Singapore-based partner that performs the actual facial matching. Any assessment of where your biometric-adjacent data goes has to account for both entities, and the SLA explicitly carves this component out.
Uploaded images are kept for 24 hours from the last interaction and only as a search-result link for sharing, and the company states that user uploads are not used for model training. But the 24-hour retention promise carries one exception worth knowing before you use the feature: images are stored permanently once you create an alert, which the company attributes to the nature of that feature. Setting an alert is therefore a materially different privacy decision from running a search.
A DMCA request removes the image from Lenso's index only and not from the site where it appeared, so getting a photo off the wider web still means contacting each website owner yourself. The DMCA Assist feature exists precisely because that second step is unavoidable.
The terms reveal that the real base API quota is 1,000 calls per month rather than the 5,000 shown on the pricing page; reaching 5,000 requires submitting a request form, refusal carries no refund, and each call returns 20 images. Anyone budgeting an integration against the advertised 5,000 should plan around 1,000 until the increase is granted.
On the review side the picture is broadly positive: on Trustpilot the product holds 4.4 rated Excellent across 168 reviews, 156 of them in the last 12 months, with 72 percent at five stars and 8 percent at one, and the platform noting that the company has replied to 38 percent of negative reviews. With 168 reviews, that sample is large enough to carry weight rather than merely indicate direction.
The substance behind the lower scores is specific and useful. The most substantive criticism concerns consistency and filtering: one reviewer described weeks of daily new hits followed by weeks of nothing and then hundreds at once, said description-based filters do not let you search for what you were wearing, and noted that thumbnails show only the face at low resolution, making it hard to tell which image is in use. More broadly, Trustpilot's own summary of the reviews records that outcomes can occasionally feel random, confusing or disappointing against expectations, and that some useful features remain behind a paywall.
Biometric Update framed the whole category as highly controversial, noting that the most prominent such tool, PimEyes, has been used by UK law enforcement and the Australian Federal Police, and that concerns about unlawful surveillance and stalking have arisen in some regions. That framing applies to any face search engine, this one included, and it is the honest context in which to evaluate the tool.
The privacy protection page claims compliance with all regional and international data privacy regulations, but points to no specific certification or third-party audit to substantiate it. There is no ISO 27001, SOC 2 or equivalent on the public pages.
Beyond the Biometric Update trade report, we found no independent product review of Lenso.ai from major outlets such as TechCrunch, The Verge, Wired or VentureBeat, and no listings on G2 or Capterra. For a product handling face data, the absence of independent scrutiny is itself worth weighing.
This section matters more here than for almost any other category of tool, so it is worth reading the commitments precisely rather than in summary.
On ownership and training: uploading transfers no rights — the company says it will not publish or reproduce your file, does not copy it into a database, and its algorithms do not learn from it, the sole purpose being comparison against the index. That is a clearer statement than many competitors offer.
On consent: face search requires explicit, personal and voluntary consent that can be withdrawn at any time, and declining any of the required consents blocks the feature entirely; the company also states it does not sell your photo and will not use it in machine learning. Consent as a hard gate, rather than a pre-ticked box, is the correct design.
On minors: the service is not intended for anyone under 16, and the company states it neither collects nor permits searches for information about people under 18, saying it uses technical measures to detect and continuously monitor such searches. That is a stronger stated position than a simple age gate.
On removal — the part that matters if you are the subject rather than the user: there are two removal routes with published turnaround: an opt-out form where the company says a single clear photo of your face suffices to remove all your pictures, and a DMCA form, with requests reviewed every business day, removals usually taking under 24 hours and DMCA requests processed immediately. Publishing a turnaround commitment at all puts this ahead of much of the field.
The honest summary: the written commitments are specific and comparatively strong, but they remain self-declared, they are split across two companies, and the alert exception quietly reverses the retention promise for anyone who uses monitoring.
There is a genuine free tier, with deliberate friction. It works with or without an account, is restricted to personal and non-commercial use, returns a portion of its results in a blocked state, and may show ads. You can see how many matches exist without paying; revealing where they are is what costs money.
At the time of writing, Starter was $19.99 per month, Professional $69.99 per month and Developer $2,800.00 per month, all plus tax, with a yearly billing option advertised at 20 percent off. The tiers differ by quota — 50, 500 and 5,000 unlocks — rather than by search quality.
Because it is switched off in your region by design. The People category is geographically restricted and is not offered where such services are not legally permitted, with the policy naming EEA countries, the UK, Argentina and the US state of Illinois.
Not Lenso. The privacy policy states plainly that the controller performs no facial search and that it is performed entirely by EYEMATCH AI PTE. LTD, a Singapore company to which your photo is forwarded. Biometric Update independently reported the partnership behind this arrangement in September 2024.
Normally 24 hours from your last interaction, held only as a search-result link for sharing, and not used for model training. The exception is important: images are stored permanently once you create an alert, which the company attributes to the nature of that feature.
Yes, and the process is documented. There are two removal routes with published turnaround: an opt-out form where the company says a single clear photo of your face suffices to remove all your pictures, and a DMCA form, with requests reviewed every business day, removals usually taking under 24 hours and DMCA requests processed immediately.
No, and this is the most common misunderstanding about the feature. A DMCA request removes the image from Lenso's index only and not from the site where it appeared, so getting a photo off the wider web still means contacting each website owner yourself.
Only within a very narrow window. Within 14 days of payment, if you have not used a single unlock and have no active alert, you can cancel from the user panel — but using one unlock or setting one alert forfeits that right. Upgrades are non-refundable even when triggered by a mistaken click, and cancelling does not return unused balance.
Fewer than the pricing page suggests at first. The base quota is 1,000 calls per month; reaching the advertised 5,000 requires submitting a request form that can be refused without a refund, and each call returns 20 images. Budget your integration against 1,000 until an increase is granted.
Broadly yes, with specific complaints. On Trustpilot it holds 4.4 rated Excellent across 168 reviews, 156 of them in the last 12 months, with 72 percent at five stars and 8 percent at one. The recurring criticisms concern inconsistent hit rates over time, filters that cannot search by clothing or activity, and low-resolution face-only thumbnails that make matches hard to identify.
