"How Should I Search?" Support for Hesitation Through Conversation. PIXTA to Publicly Release Search Experience on a Trial Basis

PIXTA Inc. (Shibuya-ku, Tokyo; President and CEO: Daisuke Furumata; TSE Standard: 3416), operator of the stock photo, illustration, video, and music material marketplace "PIXTA (Pixta)", will begin a trial release of "Chat to Search (β)", an interactive search function that evolves the existing natural language search function "Search by Text (β)", starting June 24, 2026. This will enable a search experience where users can first see search results and then gradually approach their desired materials by organizing conditions together through conversation with AI.

Background

PIXTA has offered "Search by Text (β)" since 2024 to meet visual needs that are difficult to articulate, such as when the intended use and target audience are decided, but a concrete visual image is not yet formed, and users cannot think of search keywords. With the spread of generative AI, the experience of conveying intent through text to obtain content has become widespread among users. However, conventional search functions were designed on the premise that users would verbalize what they want themselves. While images like "a bright business scene" or "a friendly family photo" exist in their minds, converting them into words that can be entered into a search box is not easy. Usage data also showed many cases where users were hesitant, thinking "I don't know how to search," and repeatedly tried again while changing search terms. In light of this situation, it became clear that converting the image in one's mind into words itself is a significant burden for users. Therefore, we have developed "Chat to Search (β)" as a new search experience that allows users to organize search conditions through conversation with AI.

Overview of "Chat to Search (β)"

In "Search by Text (β)" launched in 2024, AI displayed search results all at once based on the input. "Chat to Search (β)" goes a step further, visualizing the input content as tags for search conditions and evolving into a system where users can organize conditions while viewing search results through interaction with AI (additional questions, selection of search suggestions).

<How to Use>

User inputs ambiguous terms (e.g., "gentle spring")

Along with displaying search results, AI presents additional questions as hints for narrowing down the search and further detailed filtering options such as the color, composition, and motif of the desired material as tags.

Click on tags that match the image or enter additional keywords.

Conditions are updated, and results are narrowed down.

Continue the conversation as needed to gradually approach the desired material image.

<Evolution of Search Experience>

With conventional keyword search, users had to rewrite the search form themselves multiple times. However, with this function, conditions are narrowed down simply by continuing the conversation, allowing users to reach their target materials. Instead of users guessing the "correct input," AI organizes the user's intent together. This is a shift from "searching" to "approaching."

PIXTA search can be chosen from two methods: keyword search and "Chat to Search (β)." When clear keywords are available, use conventional keyword search. When the image is ambiguous, search for materials through conversation. Users can choose which method to use depending on their situation and objectives.

* "Chat to Search (β)" is a search function to make it easier to find materials that meet your needs from the existing digital materials handled by PIXTA.

Developer's Thoughts

Kenichi Miyamae, Head of Product Planning Department, PIXTA Business Division, PIXTA Inc.

In search, those who can convert the image in their mind into "correct words" can reach what they want faster—there has been a difference in search based on proficiency in verbalization. Now that "conveying intent with words" is becoming the norm with generative AI, we want to explore a search method that allows users to get closer to what they want through conversation, even if they cannot express it well. From search that relies on "correct input" to search that gets closer together by understanding the user's intent. "Chat to Search (β)" is an experimental initiative that represents the first step.

In the future, we hope to connect unmet needs that were not put into words with materials that were previously undiscovered, creating new encounters.

Future Plans

"Chat to Search (β)" is being released as an experimental initiative to explore new forms of search experience. We will continue to improve the accuracy of conversations and usability while observing user activity.

About PIXTA

"PIXTA" started in May 2006 as a digital material marketplace where anyone, professional or amateur, can buy and sell digital materials they have created online. It is the largest digital material site in Japan, described as "PIXTA for images and video materials related to Japanese people and culture."

In recent years, in addition to providing materials for creative production fields such as advertising, PIXTA has also been providing image and video materials for AI development as learning data for companies and academic institutions conducting machine learning. They also support AI development along with annotation services.

Furthermore, as a custom photo shooting service, they offer "PIXTA On-Demand," a business-oriented on-location shooting service that can match clients with photographers nationwide, and "PIXTA Custom," which handles all aspects of shooting, including model and location arrangements and shooting direction, responding to various visual needs of companies as a visual platform.

◆ PIXTA

https://pixta.jp

◆ PIXTA Machine Learning Data Provision

https://pixta.jp/machinelearning-dataset

◆ PIXTA On-Demand

https://od.pixta.jp/

◆ PIXTA Custom

https://pixta.jp/custom/

【Company Profile】

Company Name: PIXTA Inc. (TSE Standard: 3416)

Established: August 25, 2005

Location: JustCo Shibuya Hikarie, 33F, Shibuya Hikarie, 2-21-1 Shibuya, Shibuya-ku, Tokyo

TEL: 03-5774-2692

Capital: 332,437 thousand yen (as of the end of March 2026)

President and CEO: Daisuke Furumata

URL: https://pixta.co.jp/

Business Activities: Content Sales Business

Shooting Business

Manufacturing Experience Business

Subsidiaries: PIXTA ASIA PTE. LTD.

PIXTA VIETNAM CO., LTD.

POTONOW CO.,LTD.

YASUMI WORKS CO., LTD.