# "Ask Pocket" has different results of App vs. Web **URL:** \ **Category:** General\ **Created:** [October 18, 2026, 3:23pm UTC](https://community.heypocket.com/t/ask-pocket-has-different-results-of-app-vs-web/2303 "2026-10-18T15:23:38Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![Morty](https://yyz1.discourse-cdn.com/flex029/user_avatar/community.heypocket.com/morty/32/1885_2.png) [@Morty](https://community.heypocket.com/u/Morty)\ **Post date:** [October 18, 2026, 3:23pm UTC](https://community.heypocket.com/t/ask-pocket-has-different-results-of-app-vs-web/2303/1 "2026-10-18T15:23:39Z")
Curious if there is a setting I need to adjust in the PocketApp on my PC so that it provides the same results, regardless of using AskPocket via the website app.hpkt versus the App on my PC. I have found that the website follows my rules better than the App. I load a few reports at a time, for this presentation, I am doing just 1. (I edited names with \*\*\* for this Topic-post) **My rule/request:** _In the first sentence of the summary, state that the inspector “contacted me upon arrival to the site”. In the second sentence, state “As reported,” then continue with summarizing each of these reports into paragraphs for each report. Identify the inspectors as a Special Inspector in front of their name. Break each paragraph up by subtitles from the date of each report._ **App on PC Results:** Special Inspector M\*\*\* evaluated the installation of five 3/8" x 3-3/4" Hilti KB-TZ2 carbon steel expansion anchors in the basement-level Chiller Room. He confirmed the 3-1/2" embedment, proper edge distances, and hole cleanliness. He executed torque testing on all five anchors to 30 ft.-lbs. utilizing a calibrated wrench, yielding zero failures. The installation aligns strictly with NMA 23 RFI #22, Structural Detail 14/S5.02, and ESR# 4266. Following coordination with J\*\* from \*\*\*, he noted the overall scope remains in progress due to dimensional variances and missing clips, though all work inspected during this visit is fully compliant with approved construction documents. **App.heypocketDotCom results:** (The part bolded is just me pointing out how it followed my rule) Special Inspector M\*\*\* **contacted me upon arrival to the site. As reported,** Special Inspector M met with J\*\*\* from \*\*\* to review the approved construction documents, including NMA 23 RFI #22 and Structural Detail 14/S5.02, for anchoring duct drop supports in the basement level Chiller Room. He observed the installation of five (5) 3/8" x 3-3/4" Hilti KB-TZ2 carbon steel expansion anchors, verifying proper hole cleanliness and the required 3-1/2" embedment. Following the installation, Special Inspector M\*\*\* performed torque testing on all five anchors to 30 ft.-lbs. using a calibrated torque wrench, resulting in zero failures. While the overall scope of work remains in progress due to dimensional issues and missing clips at certain locations, the installed work was inspected and found to be in full compliance with the approved HCAI construction documents and ESR# 4266. I checked my Settings, both are using Gemini 3 Flash Preview. Both I am choosing the “PRO” option to summarize. Yet the results are different. For this presentation for this post, I ran both processes within about a second apart from each. As long as it took to move my mouse from left side of screen to the right as I had them both queued up. Other little things that are different are - How it identifies quantities- five vs (5) - Notes the inspector “observed” vs “executed” the torque - DotCom did a better job noting what they were actually anchoring that day (duct drop supports). PC never mentioned it. - DotCom notes how they met with J in the beginning, while PCApp mentions J\* later in the results. I like the website App.HPdotcom better as it does what I ask but would prefer to use the PC app as I tend to accidentally close the browser when I’m clicking around with documents. Any ideas? (Just for curiosity, I ran the test again after typing everything above, same results.) ---
**Author:** ![Vic](https://yyz1.discourse-cdn.com/flex029/user_avatar/community.heypocket.com/vic/32/2532_2.png) [@Vic](https://community.heypocket.com/u/Vic)\ **Post date:** [October 19, 2026, 11:32am UTC](https://community.heypocket.com/t/ask-pocket-has-different-results-of-app-vs-web/2303/2 "2026-10-19T11:32:48Z")
Thanks for sharing such a detailed comparison, Morty. There isn’t a Desktop setting you need to change. Ask Pocket generates each response independently, so wording, ordering, and number formatting can vary between runs, even with the same PDF, prompt, and Pro option. We also checked the matching requests and found that the Desktop response included the requested “contacted me upon arrival…” and “As reported…” opening immediately before the dated section, though that portion isn’t included in the Desktop result pasted here. If those sentences were missing in the Desktop app itself, could you share a screenshot and your Pocket Desktop version? That would help us check whether the app displayed or copied only part of the response