Seventeen of Pensive's eighteen public pages are served with the identical title "Pensive Grader" and the same one-line description: the AI tutor, the plans, the about page, all five case studies and both blog posts. Search engines and AI assistants see eighteen copies of one page, so "AI tutor for my course", "AI grading handwritten exams", "UC Berkeley Data 100 AI grading" or "AI grading accuracy" have nothing specific to match. The engineering page, the one exception, shows the fix already works in the site's code.
First action
Write titles and descriptions for the 17 pages in one sheet.
How to measure it
Search impressions for non-brand queries
- Now
- –
- Target
- 3x
- By
- 90 days
Steps
List pages
All 18 sitemap URLs with the search each should win.
Growth lead 2 hours
Write tags
Title under 60 characters and description under 155 for each.
Growth lead 1 day
Ship server-side
Same mechanism as the engineering page.
Founding Engineer 2 days
Social previews
Open Graph text and images per page.
Product Designer 2 days
Request indexing
Search Console for all 18 URLs.
Growth lead 1 hour
Track queries
Impressions and clicks per page.
Growth lead weekly
Template rule
Every new page needs its own tags before launch.
Founding Engineer 1 hour
Why it works
Use the engineering page's pattern on every page: a title and description written for the search each page should win ("AI Tutor Built for Your Course | Pensive", "UC Berkeley Data 100 cut TA grading time 60% | Pensive case study", "How Pensive reaches 98% accuracy on high-confidence grades"), unique social preview text, and the same tags in the server-rendered HTML rather than only after the app loads. Then check Search Console for which queries each page starts to win.
Instructors research tools by searching their problem and their peers' results. Distinct titles let each case study and the tutor page rank and be quoted on their own, at almost no cost, and every later content and department page inherits the fix.
Ready to use
ChecklistPer-page tag sheet
[ ] /schools: AI grader for handwritten exams and problem sets [ ] /schools/tutor: AI tutor built for your course [ ] /schools/plans: Pensive plans: bubble sheet, Lite, Pro, Max [ ] Each case study: institution, course and result [ ] Each blog post: its own headline [ ] Unique description with one number per page [ ] Same tags in the served HTML [ ] Social preview image per page
Risks and how to handle them
Titles over-promise
Use the numbers from each case study exactly.
Client-side tags overwrite server ones
Check the rendered and served HTML both match.
Evidence
- Your websiteSitemap (18 URLs, all under /schools, checked October 23, 2026): 17 of the 18 pages are served with the same title, "Pensive Grader", and the same description ("AI Grading Assistant that reduces grading time, generates rubrics, and give students timely feedback."), including the tutor page, plans, about, the five case studies and both blog posts. Only /schools/engineering has its own title ("Pensive for Engineering Departments | AI Grader") and description.Source
- Your websiteEngineering department page: "Give your engineering faculty their weekends back"; names the problems ("Many graders, many readings. The same rubric drifts across TAs and sections"; "Feedback that arrives too late"); a video story of Dr. Yan Tang at Embry-Riddle whose graders "finished their work without weekend grading"; proof from UC Berkeley ("40,000+ Submissions processed"; "60% Reduction in TA grading time"); call to action "Book a 20-minute demo". It is the only department page.Source
- Your case studiesUC Berkeley Data 100 case study: 800-1,200 students a semester and up to fifty TAs; one assignment took staff 65 hours without AI and an estimated 27 hours with Pensive, "a 60% reduction"; the case study values the 38 hours saved at "$2,280+ in instructional value" at $60 an hour for tutoring; it also compared human and Pensive grades on 100 randomly selected submissions for grading equity.Source
- Your blog"Beyond the Black Box" (February 9, 2026): a grade is labelled High Confidence only when the AI grade passes Pensive's checks and the handwriting transcription is clear; "Currently, our High Confidence accuracy rate is 98%", defined as High Confidence scores left unchanged after instructor review and regrade requests.Source
- Your websiteTutor page: "The AI Tutor Built for Your Course", with course-specific guidance "available 24/7", citations to lecture notes, AI policy control "with one click", a "Compare Pensive with ChatGPT!" demo and a research paper with UC Berkeley that shows the tutor "enhances student satisfaction, encourages collaboration, and increases question volume". Student testimonials from Berkeley courses; one reports a 5-10% rise in exam scores.Source