AI Search Pulse
Document AI / IDP Monthly Benchmark: September 2026
Benchmark Date: September 2026
Methodology Version: 1.0
AI Assistant: ChatGPT
Model: GPT-5.6 Luna
Tracked Vendors: 10
Prompt Library: 49 Buyer Evaluation Scenarios
Executive Summary
This is the third monthly benchmark published by AI Search Pulse for the Intelligent Document Processing (IDP) category, following the July 2026 baseline and August 2026 benchmark.
Using the AI Search Pulse Research Methodology (Version 1.0), we submitted the same standardized library of 49 buyer-representative prompts to ChatGPT and recorded which tracked enterprise software vendors were recommended, how frequently, and across which buyer scenarios.
September produced a more concentrated competitive picture than August. ABBYY, UiPath, and Hyperscience continued to form the leading group, with all three accounting for substantially more AI recommendation visibility than the remaining tracked vendors. ABBYY and UiPath each recorded 41 mentions, followed by Hyperscience with 39.
One important methodological difference should be noted. The August benchmark was conducted from Toronto, while the September benchmark was conducted from India. Geographic context can influence AI-generated recommendations, particularly when models retrieve or incorporate regionally relevant information. As a result, month-over-month changes should be treated as directional rather than as a controlled measurement of changes in vendor visibility.
This benchmark measures AI recommendation visibility, not product quality, customer satisfaction, analyst rankings, or market share.
Key Findings
Before the detailed metrics, four patterns stood out in this benchmark.
1. ABBYY and UiPath moved into an effective tie at the top
ABBYY and UiPath each recorded 41 tracked recommendations, giving both vendors a 20.2% share of total tracked recommendations.
Hyperscience remained close behind with 39 mentions and a 19.2% share. Together, the three vendors accounted for almost 60% of all tracked recommendations in the September benchmark.
This reinforces the concentration already visible in August, when the same three vendors occupied the top three positions.
2. Hyperscience remained highly visible across buyer scenarios
Hyperscience recorded 39 total mentions, with strong visibility across all four prompt categories.
Its presence was particularly consistent in competitive prompts, where it appeared 11 times, matching both ABBYY and UiPath. This suggests that Hyperscience continues to be surfaced not only in general enterprise recommendations but also when AI systems are explicitly comparing IDP vendors.
3. Rossum remained the strongest vendor outside the leading group
Rossum recorded 28 mentions and a 13.8% share of voice.
Its competitive visibility remained particularly strong, with 10 mentions across the 12 competitive prompts. Rossum also maintained meaningful API visibility, appearing in 7 developer/API scenarios.
The gap between Rossum and the top three remains significant, but its distribution across buyer scenarios continues to make it one of the more consistently surfaced vendors in the benchmark.
4. Geographic context is now an important consideration
The September benchmark was conducted from India, compared with Toronto for the August run.
Because AI systems can use regional search results, sources, market context, and other location-sensitive signals when generating recommendations, this change introduces another variable into the month-over-month comparison.
The September results therefore provide a useful snapshot of AI recommendation visibility, but changes from August should not automatically be interpreted as gains or losses in underlying vendor visibility.
Continued tracking from a consistent geographic location will make future month-over-month comparisons more controlled.
AI Visibility Summary
| Vendor | Total Mentions | Enterprise | Use Case / Industry | API | Competitive | AI Visibility Score |
|---|---|---|---|---|---|---|
| ABBYY | 41 | 10 | 12 | 8 | 11 | 20.20% |
| UiPath | 41 | 10 | 12 | 8 | 11 | 20.20% |
| Hyperscience | 39 | 9 | 11 | 8 | 11 | 19.21% |
| Rossum | 28 | 6 | 5 | 7 | 10 | 13.79% |
| Tungsten Automation | 18 | 7 | 4 | 3 | 4 | 8.87% |
| Nanonets | 15 | 2 | 2 | 4 | 7 | 7.39% |
| Docsumo | 6 | 1 | 1 | 1 | 3 | 2.96% |
| Klippa | 6 | 1 | 1 | 1 | 3 | 2.96% |
| Veryfi | 6 | 1 | 1 | 2 | 2 | 2.96% |
| Affinda | 3 | 1 | 1 | 1 | 0 | 1.48% |
| Total | 203 | 100.00% |
AI Visibility Score represents each vendor’s share of total tracked recommendations within the benchmark prompt library.
Benchmark Observations
The leading group became more concentrated
ABBYY, UiPath, and Hyperscience continued to dominate AI recommendations in September.
ABBYY and UiPath each recorded 41 mentions, while Hyperscience recorded 39. Their combined 121 mentions represented approximately 60% of all tracked recommendations in the benchmark.
This is broadly consistent with the leadership pattern observed in August, although the geographic change means the movement should not be interpreted as a controlled month-over-month increase in visibility.
ABBYY and UiPath showed broad category coverage
Both vendors appeared consistently across enterprise, use-case, API, and competitive prompts.
ABBYY recorded 10 enterprise mentions, 12 use-case mentions, 8 API mentions, and 11 competitive mentions. UiPath showed the same category distribution.
The breadth of this visibility is notable because it indicates that their presence is not dependent on a single buyer scenario.
Hyperscience remained competitive across the full funnel
Hyperscience recorded 9 enterprise mentions, 11 use-case mentions, 8 API mentions, and 11 competitive mentions.
Its 11 competitive mentions place it alongside ABBYY and UiPath in direct comparison scenarios. This suggests that the vendor is being surfaced not only as a general IDP recommendation but also as part of active vendor-selection conversations.
Rossum continued to perform strongly in competitive scenarios
Rossum recorded 10 competitive mentions, its highest category count.
It also appeared in 7 API scenarios and 6 enterprise scenarios. While its use-case visibility was lower at 5 mentions, its overall distribution remained relatively broad.
This continues the pattern observed in August, where Rossum showed meaningful visibility across different buyer personas rather than being concentrated exclusively in developer-oriented prompts.
Tungsten Automation remained stronger in enterprise than use-case scenarios
Tungsten Automation recorded 18 total mentions.
Its strongest category was enterprise, with 7 mentions, followed by competitive prompts with 4 and use-case prompts with 4. API visibility was comparatively limited at 3 mentions.
This creates a different visibility profile from vendors such as Rossum and Nanonets, whose recommendations were more closely tied to developer/API or competitive scenarios.
Nanonets remained particularly visible in competitive prompts
Nanonets recorded 15 total mentions, including 7 competitive recommendations.
Its enterprise and use-case visibility remained considerably lower, with 2 mentions in each category, while API visibility reached 4.
This suggests that Nanonets continues to have stronger visibility when AI systems are asked to compare or shortlist vendors than when they are asked for broad enterprise or industry-specific recommendations.
Smaller vendors remained concentrated in competitive scenarios
Docsumo and Klippa each recorded 6 mentions, while Veryfi recorded 6 and Affinda recorded 3.
For Docsumo and Klippa, three of their six mentions came from competitive prompts. Veryfi recorded two competitive mentions and two API mentions.
Affinda appeared once each in enterprise, use-case, and API scenarios but did not appear in the competitive prompt set.
For these vendors, the relatively small number of total recommendations means individual prompt results can have a meaningful effect on their overall score. Continued monthly tracking will be needed before interpreting these differences as persistent visibility patterns.
What Changed From August?
September produced a somewhat more concentrated distribution of recommendations among the leading vendors.
| Vendor | August SOV | September SOV | September Mentions |
|---|---|---|---|
| ABBYY | 18.75% | 20.20% | 41 |
| UiPath | 18.30% | 20.20% | 41 |
| Hyperscience | 16.96% | 19.21% | 39 |
| Rossum | 13.84% | 13.79% | 28 |
| Tungsten Automation | 10.71% | 8.87% | 18 |
| Nanonets | 8.04% | 7.39% | 15 |
| Docsumo | 4.02% | 2.96% | 6 |
| Klippa | 4.02% | 2.96% | 6 |
| Veryfi | 3.57% | 2.96% | 6 |
| Affinda | 1.79% | 1.48% | 3 |
The comparison is useful for identifying patterns, but it should not be treated as a pure measure of vendor-specific change because the August and September runs were conducted from different geographic locations.
The most notable pattern is the continued strengthening of the three-vendor leadership group. ABBYY, UiPath, and Hyperscience all increased their share of tracked recommendations relative to the August calculation.
At the same time, Tungsten Automation and Nanonets represented a smaller share of the September recommendation set.
Whether these movements reflect actual changes in AI visibility, geographic effects, prompt-level variation, or a combination of factors will become clearer with additional same-location benchmarks.
Questions for the Next Benchmark
- Do ABBYY, UiPath, and Hyperscience continue to form the leading group in the next benchmark?
- Does Tungsten Automation recover some of its use-case visibility?
- Does Nanonets maintain its relatively strong competitive visibility?
- Does Rossum continue to perform strongly across competitive and API scenarios?
- How much variation is attributable to geographic context when the benchmark is run from a consistent location?
- Do the same visibility patterns persist across another AI model or platform?
Methodology & Limitations
This benchmark measures AI recommendation visibility within a standardized set of buyer evaluation scenarios. It does not evaluate product quality, analyst rankings, customer satisfaction, or market share.
The September benchmark used the same 49-prompt library and the same 10 tracked IDP vendors used in the previous benchmark.
Each vendor mention was counted once per prompt based on the tracked-company results. The September benchmark produced 203 total tracked vendor recommendations across the 49 prompts.
Some prompts returned no tracked vendors or produced recommendations outside the IDP category. These were not counted toward the tracked-company scores.
A geographic difference is also present between the August and September runs. The August benchmark was conducted from Toronto, while the September benchmark was conducted from India. Geographic context can influence AI-generated recommendations, particularly when systems incorporate search results or regionally relevant information. This means month-over-month SOV changes should be treated as directional rather than as a controlled experiment.
Future benchmarks will provide a stronger basis for identifying persistent vendor-level visibility patterns as more same-method, same-location data becomes available.
AI Search Pulse Roadmap
| Status | Milestone |
|---|---|
| ✅ Published | Research Methodology v1.0 |
| ✅ Published | Document AI / IDP Baseline Benchmark (July 2026) |
| ✅ Published | Document AI / IDP Monthly Benchmark (August 2026) |
| ✅ Published | Document AI / IDP Monthly Benchmark (September 2026) |
| 📅 Planned | ChatGPT + Gemini Comparative Benchmark |
AI systems like ChatGPT, Google AI Overviews, and Perplexity are becoming an increasingly important layer in how buyers discover and evaluate software.
AI Search Pulse tracks that visibility over time, using standardized buyer scenarios to understand which vendors are being surfaced, where they appear, and how those patterns change.
