How to Use AI to Identify and Authenticate a Handmade Rug

Last Updated: June 2026

You are at an estate sale, an antique market, or scrolling through a listing online. There is a rug. It looks handmade. The seller says it is Persian. The price suggests it might be genuine. But you are not an expert, and you have no way to know for certain whether you are looking at a 19th century Kashan or a machine-made imitation from a warehouse in Guangzhou.

Until recently, your only options were to trust the seller, hire a specialist, or walk away. Now there is a third option: use AI.

Artificial intelligence tools have become surprisingly capable at analyzing rug images, identifying design patterns, suggesting regional origins, and flagging visual characteristics that help distinguish handmade pieces from machine-made imitations. They are not infallible - and we will be clear about where they fail - but used correctly alongside the physical tests that no algorithm can replace, AI has become a genuinely useful first step in evaluating any rug you are considering buying.

This guide covers every AI tool worth using for rug identification and authentication, how to get the best results from each one, what they can and cannot reliably tell you, and the physical tests you must still perform in person.

Google Lens: The Fastest Starting Point

Google Lens is the most immediately accessible AI tool for rug identification and the one most people already have in their pocket. On any smartphone, open the Google app or camera, point it at the rug, and Google Lens will analyze the image and return visual matches from across the web.

For rug identification, Google Lens works by matching visual patterns in your photograph against its index of rug images. It is particularly good at identifying well-documented rug types with distinctive visual signatures. Point it at a classic Bokhara gul pattern and it will return results showing Bokhara rugs from multiple sources. Point it at the characteristic stepped medallion of a Kazak rug and it will pull up Kazak rug references. The more distinctive and well-documented the design, the more reliable the result.

For best results with Google Lens, photograph the rug in good natural light with the camera directly overhead rather than at an angle. Capture a section that shows the main field pattern clearly rather than the border or fringe. If the full rug is large, photograph a representative section of the field in close enough detail that the individual motifs are clearly visible. Run multiple searches on different sections of the rug - the field, the main border, and any corner medallions - and compare the results.

Google Lens is less reliable for unusual, regional, or tribal pieces that are less well-documented online. A mainstream Khal Mohammadi rug will generate accurate matches. A rare Shahsavan tribal flatweave may return vague or incorrect suggestions. Use the results as a starting point for further investigation rather than a definitive identification.

ChatGPT and Claude: Deep Pattern Analysis

Large language model AI assistants - ChatGPT with vision, Claude, and similar tools - offer a different kind of analysis from Google Lens. Rather than visual matching, they apply pattern recognition combined with trained knowledge about rug history, design vocabulary, and regional traditions. Upload an image and ask specific questions, and a capable AI assistant will walk you through what it can observe.

The most useful questions to ask an AI assistant when analyzing a rug image include:

What design tradition does this pattern suggest? Can you identify specific motifs and describe their likely regional origin? What weaving tradition does the border system suggest? Does the color palette indicate natural or synthetic dyes? What does the regularity or irregularity of the pattern suggest about the construction method?

A well-prompted AI assistant will often identify specific design elements with impressive accuracy. It may recognize a Herati pattern in the field of a Persian rug, identify the fil pai motif characteristic of Khal Mohammadi production, distinguish between the gul formats of different Turkmen tribal traditions, or note that a floral design shares characteristics with the Tabriz weaving school. This kind of analysis would have required a specialist a decade ago.

For authentication purposes, ask the AI to comment specifically on pattern regularity. Genuine hand-knotted rugs show subtle irregularities in their pattern - slight variations in motif size, minor deviations in straight lines, the slight inconsistency that comes from human hands working from memory or a coded design chart rather than a mechanical process. Machine-made rugs are mathematically perfect. Ask the AI: does the pattern show the subtle irregularities consistent with hand production, or does it appear mechanically uniform?

ChatGPT with the vision model enabled, Claude, and Google's Gemini all perform well at this kind of analysis. The quality of the result depends heavily on the quality of the photograph and the specificity of your questions. Vague questions get vague answers. Specific, well-framed questions about particular design elements get genuinely useful responses.

Reverse Image Search: Finding Provenance

Standard reverse image search - through Google Images, TinEye, or Bing Visual Search - serves a different purpose from pattern identification. It tells you where an image has appeared before online, which can reveal important provenance information or raise red flags.

If you are evaluating a rug from an online listing, save the seller's photographs and run them through reverse image search. If the same rug images appear on multiple different seller websites at different prices, or if the images appear on a stock photo site or a manufacturer's catalog, you are looking at a misrepresented listing. Legitimate handmade rug sellers photograph their actual inventory - each piece is unique and one-of-a-kind, so the same photograph should not appear under multiple different ownership claims.

This technique is particularly useful when evaluating rugs on marketplace platforms like eBay, Etsy, or similar sites where misrepresentation is more common than in specialist rug dealerships.

AI for Pile Analysis: What to Look For

Beyond design identification, AI image tools can assist with analyzing the pile structure visible in close-up photographs, which is one of the most reliable indicators of construction method.

Photograph the surface of the rug in close-up using your smartphone's macro mode or by zooming in to maximum detail. The pile structure of a genuine hand-knotted rug looks organic and slightly irregular when examined closely. Individual tufts emerge from the foundation in a way that varies slightly in density and height across the surface. The transition between colors in the pattern shows slight feathering at the edges where adjacent knots of different colors meet.

Machine-made rugs have a pile structure that looks mechanical and uniform under close examination. The tufts are perfectly even, perfectly spaced, and the color transitions between design elements are crisp and abrupt in a way that genuine hand-knotted production cannot achieve.

Upload a close-up pile photograph to an AI assistant and ask specifically: does the pile structure appear consistent with hand-knotting or mechanical production? Does the color transition between design elements show the slight irregularity consistent with individual hand-tied knots? The answer will not be definitive but it adds another data point to your evaluation.

The Limits of AI Authentication

Honest disclosure: AI tools are a useful first filter, not a final verdict. There are specific situations where AI analysis is unreliable and where you should not rely on it.

AI cannot assess texture, weight, or flexibility - all of which are critical indicators of construction quality and material authenticity. A genuine hand-knotted Afghan rug has a characteristic weight and firmness that no photograph can convey. A Bidjar rug is famously stiff and heavy in a way that immediately distinguishes it from any other rug type - but that distinction is entirely physical and entirely invisible to an AI analyzing an image.

AI cannot perform the burn test for fiber identification. The single most reliable test for distinguishing genuine silk from mercerized cotton or viscose imitations is to pull a few fibers and hold them to a flame. Genuine silk burns like hair, smells like burning protein, and leaves a crushable dark ash. Artificial silk burns like paper. No AI can smell or feel - the burn test must be done in person.

AI cannot assess the back of the rug in the detail required for reliable authentication. The reverse of a genuine hand-knotted rug shows individual knots in a pattern that closely mirrors the front - a level of detail that requires physical examination or an extremely high-resolution close-up photograph taken under controlled lighting. Most rug listing photographs do not show the back at all, and those that do rarely show sufficient detail for reliable AI analysis.

AI can be fooled by high-quality machine-made rugs designed to look handmade. Some machine production is now sophisticated enough that surface photographs are difficult to distinguish from genuine hand-knotted pieces. The physical tests described below remain the only reliable authentication method for borderline cases.

For a complete guide to the physical authentication tests that AI cannot replace, see our post on how to tell if a rug is handmade.

A Practical Authentication Workflow

Here is a step-by-step approach that combines AI tools with physical examination for the most reliable result:

Start with Google Lens on the full rug to get an initial design identification and suggested origin. Note the rug type suggested and whether the results are consistent and specific or vague and inconsistent.

Run a reverse image search on any seller photographs to check for misrepresentation red flags.

Upload the image to ChatGPT or another AI assistant with vision capability. Ask specific questions about the design vocabulary, motif identification, regional characteristics, and pattern regularity. Note any observations about irregularities consistent with hand production.

Photograph the pile in close-up and ask the AI to assess the pile structure for hand versus machine construction indicators.

Then move to physical examination. Turn the rug over and examine the back for individual knots. Check the fringe for structural versus decorative attachment. Part the pile to examine the knot base. Perform the fold test. If authenticity of fiber type matters, perform the burn test on a few pulled threads.

Cross-reference your AI findings with the physical examination. Consistent findings across both reinforce each other. Contradictions are a signal to examine more carefully before committing to a purchase.

Using AI to Identify Specific Rug Types

Here is a quick reference for what AI tools do best when identifying the major rug types available at ALRUG:

Bokhara rugs are among the easiest for AI to identify correctly due to their highly distinctive gul medallion format. Google Lens is reliable here.

Khal Mohammadi rugs are similarly distinctive with their dense burgundy fields and fil pai patterning. AI identification is generally reliable.

Kazak rugs have a distinctive bold geometric vocabulary that AI recognizes well, though the many regional variations within the Kazak family can cause confusion.

Oushak rugs are identifiable by their characteristic large-scale floral medallions and muted ivory and gold palette. AI performs well here.

Gabbeh rugs are distinctive enough - large color fields, naive figurative elements - that AI identification is generally reliable, though distinguishing genuine Qashqai production from commercial imitations requires physical examination.

Kilim rugs are identifiable by their flat, reversible surface and bold geometric flat-weave patterns, but distinguishing Afghan vegetable kilims from Turkish or Iranian production requires close examination of construction details that may not be visible in photographs.

Baluchi rugs with their dark jewel-tone palette and complex geometric fields are identifiable by AI, though the full range of the Baluchi tradition is less well-documented online than commercial rug types.

Ziegler and Chobi rugs with their washed-out floral palette are generally identifiable, though they can be confused with other washed or abrash-effect rug types.

Overdyed rugs are a category where AI identification is particularly useful - the combination of vintage hand-knotted foundation and bold contemporary overdyed color is quite distinctive visually.

Persian rugs from specific city traditions - Tabriz, Kashan, Isfahan, Nain - are identifiable to varying degrees. The major city traditions are well-documented online and AI identification is reasonable. Rarer village and tribal types are less reliably identified.

Vintage rugs present particular challenges for AI because age-related characteristics - abrash, color mellowing, pile wear - can be difficult to distinguish from artificially distressed production in photographs alone.

Afghan war rugs are among the most distinctive and recognizable rug types in the world - their imagery of tanks, weapons, and military equipment is unmistakable - and AI identification is highly reliable.

What Good AI Results Look Like

A reliable AI identification result has several characteristics. It is specific rather than generic - it suggests a particular weaving tradition, not just "oriental rug." It is consistent across multiple tools and multiple photographs of the same piece. It matches what you can observe yourself about the design vocabulary and color palette. And it is accompanied by hedging language that acknowledges uncertainty rather than false confidence.

An unreliable result is one that gives different answers from different photographs of the same rug, produces vague generic responses like "hand-woven carpet from Central Asia," or gives highly specific confident answers for a rug type that is unusual or rare enough that training data would be limited.

Use AI identification as a well-informed starting hypothesis, not a conclusion. The physical examination is what confirms or contradicts it.

Building Your Knowledge

The most reliable way to identify and authenticate handmade rugs over time is to develop your own knowledge of the major weaving traditions - what each type looks like, what its construction characteristics are, and what distinguishes a quality example from a lesser one. AI tools accelerate this process by helping you understand what you are looking at as you encounter new pieces.

Our blog covers the major handmade rug traditions in depth. If the AI suggests your rug might be a Gabbeh, a Bokhara, a Baluchi, a Bakhtiari, a Bidjar, or a Khal Mohammadi, follow the link and read the full guide. Understanding the tradition behind the type makes you a far more effective evaluator than any AI tool alone.

Our entire collection at ALRUG consists of verified genuine hand-knotted rugs sourced directly from weavers in Afghanistan and Pakistan. Every piece passes the physical tests described in this guide. You can explore the full range at all rugs or shop by style, size, or color using the navigation above. If you need a specific piece that is not in our current stock, our custom rug service produces any hand-knotted design to your exact specifications with free worldwide shipping.

Frequently Asked Questions

Can AI really identify a handmade rug from a photograph? Yes, with important limitations. AI tools like Google Lens and ChatGPT with vision capability can identify well-documented rug types with reasonable accuracy, particularly for mainstream commercial types like Bokhara, Oushak, and Kazak rugs. They are less reliable for unusual, rare, or regional pieces. AI identification should always be confirmed with physical examination before making a purchase decision.

What is the best AI tool for identifying a rug pattern? Google Lens is the fastest starting point for visual matching against known rug types. ChatGPT or Claude with vision capability provides deeper analytical commentary about design vocabulary, regional characteristics, and construction indicators. Using both together gives the most complete picture.

Can AI detect a fake rug? AI can identify visual characteristics consistent with machine production - perfect pattern regularity, uniform pile structure, mechanical color transitions - that suggest a rug may not be genuinely hand-knotted. But AI cannot perform the physical tests that are the most reliable authentication methods: examining the back for individual knots, checking the fringe structure, performing the burn test for fiber identification, and assessing weight and flexibility. Use AI as a first filter and physical examination as the final verification.

What should I photograph to get the best AI rug identification result? Photograph the full rug from directly overhead in good natural light. Then photograph the main field pattern in close-up detail. Photograph a border section. Photograph the back of the rug to show the knot structure. The more detailed and well-lit your photographs, the more useful the AI analysis will be.

Is Google Lens accurate for rug identification? Google Lens is reasonably accurate for well-documented mainstream rug types with distinctive visual signatures. It is less reliable for tribal, regional, or unusual pieces that are less well-represented in its training data. Use it as a starting point and cross-reference results with other AI tools and physical examination.

What physical tests should I always perform regardless of what AI says? Turn the rug over and examine the back for individual knots. Check whether the fringe is structural or sewn on. Part the pile to see the knot base. Perform the fold test. For fiber verification, perform the burn test on pulled threads. These physical tests are described in full in our guide on how to tell if a rug is handmade.

Can AI identify the age of a rug? AI can identify visual characteristics associated with age - color mellowing, abrash effects from natural dyes, pile wear patterns - but cannot reliably date a rug from photographs alone. Artificial distressing and intentional abrash effects in contemporary production can mimic age convincingly in images. Physical examination by a specialist is required for reliable dating of antique pieces.