AP Computer Science Principles rewards students who can read a working program and explain what it does in plain English. The Explore Performance Task is the part of the AP CSP exam where that skill is graded under a microscope, and most students who walk out of a 5 think they lost a row they had never heard of. The task is short on the calendar (roughly eight classroom hours) but heavy on the rubric. If you want a 5, you have to understand the five rows the College Board readers actually mark, not the marketing description of the task.
Before unpacking those rows, a quick frame: AP CSP has two through-course performance tasks (Explore and Create) submitted digitally, plus a multiple-choice end-of-course exam. The Explore PT contributes about 15% of the composite AP score, the Create PT another 15%, and the multiple-choice exam 70%. The Explore PT is a written response to a computing innovation the student has not built, drawn from a category list released by the College Board. This article centres on how the written response is scored, where candidates typically drop rows, and what to do about it.
What the Explore PT actually asks you to produce
For the Explore PT, students pick one computing innovation from a College Board-provided list, then write a written response that includes a computational artifact (a screenshot, video, or other visual) and an explanation of how the innovation works, its purpose, and its effect. The artifact is not code the student wrote; it is something that shows the innovation in action. The written response is the graded component, and the task is scored on a small number of clearly defined rubric rows.
Candidates work on the task in class and submit it through the College Board's digital portfolio. There is no in-class time pressure to finish in one sitting, but teachers typically pace it across two to three weeks. The submission must include a reference to a computing innovation, a working link or screenshot, and a written response of around 300–500 words in the College Board's template.
The five rubric rows readers actually mark
The College Board's Explore PT scoring guideline uses five rows. I will name them the way the published rubric does, and then walk through how each one is awarded in practice.
- Row 1: Identifying a computing innovation. The response names a specific innovation and correctly categorises it (for example, "self-driving car navigation systems," not "technology").
- Row 2: Explaining how the innovation uses data. The response describes what kind of data the innovation collects, processes, or stores, and how that data moves through the system.
- Row 3: Explaining the purpose and effect of the innovation. The response connects the innovation to a real-world audience and identifies a meaningful effect, beneficial or harmful.
- Row 4: Describing how the innovation functions, including an algorithm. The response explains the logic of how the innovation works at a level a non-expert can follow, including at least one identifiable algorithm or step-by-step process.
- Row 5: Identifying and explaining a beneficial and a harmful effect, with a credible source. The response names a specific benefit and a specific harm, each supported by a cited source of appropriate quality.
Each row is awarded as 0 or 1, which gives a raw Explore PT score of 0–5. That raw score is converted into the same 1–5 AP scale used across the exam, and then weighted into the final composite.
Where students quietly lose Row 4
Row 4 is the algorithm row, and it is the one where the largest share of 4s drop a point. Students read "explain how it works" and write a paragraph about what the innovation does for the user. That is Row 3 territory. Row 4 wants the actual mechanism: what input is taken, what transformation happens, what output is produced, and what sequence of steps carries that out.
For a self-driving car, a Row 4 sentence sounds like: "The vehicle's perception system takes a stream of LiDAR and camera input, runs an object-detection algorithm to label each object in the frame, and feeds those labels into a planning module that produces a short-horizon driving trajectory." That sentence names input, transformation, and output. A Row 4 that loses the point usually has only the first and the last. The middle is left to the reader's imagination, and the reader will not fill it in for you.
A practical test: if you delete every noun in the algorithm sentence and the sentence still reads sensibly, the row is too generic. If you cannot replace the innovation with a different one without rewriting the sentence, the row is specific enough to earn the point.
Why Row 5 trips even strong writers
Row 5 asks for one beneficial effect and one harmful effect, each backed by a credible source. The point is usually not lost on the benefits side. It is lost in one of three places: the harm is generic ("could be misused"), the source is not actually a source (a URL with no author or publication), or the response treats the effect as a sentence at the end of the document instead of inside the running explanation.
For most candidates I work with, the fix is to choose a harm that has been reported in a named outlet (a newspaper, a peer-reviewed paper, a court filing, an FTC report) and to integrate the citation in the body of the paragraph, not in a separate "Sources" list appended at the end. Readers will mark the row only if they can see the source do work in the argument.
"The Amazon Echo's always-on microphone has been documented in multiple court records to have captured and transmitted private conversations without the wake word being spoken," is a Row 5 sentence. "Privacy could be a concern" is not.
Common pitfalls and how to avoid them
Across several years of marking these responses against the published rubric, four pitfalls show up often enough to be worth naming directly.
- Choosing an innovation that is too broad. "Social media" is not an innovation. "The Instagram Explore recommendation algorithm" is. If you can fit your answer on a T-shirt, the reader will mark Row 1 against you.
- Describing a feature instead of an algorithm. Row 4 wants a mechanism, not a user-facing description. "The app shows you videos" is a feature. "The app scores each video on watch-time and similarity and serves the top 20 in a ranked feed" is an algorithm.
- Using the same source twice. If the benefit and the harm are both supported by the same blog post, the reader has no way to tell that you actually surveyed the effect. Two distinct sources, one for each side, is the safer pattern.
- Submitting the artifact without a caption. The artifact carries weight only if the response tells the reader what to look at. A one-line caption tying the artifact to Row 4 is the cheapest point on the rubric.
How the Explore PT score becomes an AP score
The five-row raw score is converted to a 1–5 AP score using a published scale that has shifted in recent administrations. Because the conversion table is updated each year, treat any specific number-to-AP mapping you see online as illustrative, not authoritative, and verify against the current College Board scoring document. The end-of-course exam is scored separately on a multiple-choice scale, and the two performance tasks are weighted in to produce the composite. A simple comparison of weighting is below.
| Component | Weight in composite | Format |
|---|---|---|
| Multiple-choice exam | 70% | ~70 questions, 2 hours |
| Create Performance Task | 15% | Written response + code of own program |
| Explore Performance Task | 15% | Written response about a chosen innovation |
The practical implication: a perfect Explore PT is not enough on its own. The task is one of three legs of a composite score, and the multiple-choice exam still does the heavy lifting. Treat the Explore PT as the row of points you can guarantee, not the row that carries a 5 on its own.
How to study for the Explore PT without writing ten drafts
Most students do not need ten full drafts. They need two carefully marked drafts and a checklist. The checklist I have students build has one line per rubric row, and a student cannot move from draft one to draft two until every line has been checked by another reader.
- Draft 1 with the chosen innovation in place. Hand it to a peer or a teacher with the five-row rubric printed above it. Mark every row 0 or 1 before the student leaves the room.
- Rewrite only the rows that lost the point. Do not rewrite the rows that earned it; polishing an earned row spends time you do not have.
- Run the final draft past the same reader a second time. If the row score matches the first read, submit. If it does not, the response is not finished.
- Cite every source inline, in the body of the paragraph, with author and publication. A "Sources" list at the end is not a substitute.
- Submit at least 24 hours before the deadline. The College Board's digital portfolio has a habit of queuing uploads near the cutoff, and a last-minute submission that fails to sync is not a row you can earn back.
For most candidates I work with, this two-draft loop catches the four pitfalls above before the response ever leaves the room. The single biggest change is moving the source citation from the end of the document into the body of the paragraph; the second is rewriting the algorithm sentence until it has a verb that is not "uses."
What to do next if Explore is the leg you are weakest on
If your Create PT is already in good shape and your multiple-choice practice is steady, the Explore PT is the highest-leverage 15% of the composite. Pick an innovation you can describe in two sentences to a non-technical friend, build a one-paragraph outline against the five rows, and write the first draft against the rubric before you touch a word of the final response. The rows will tell you which paragraphs need rewriting.
AP Courses' one-to-one AP Computer Science Principles programme scores every Explore draft against the current College Board rubric and rewrites the algorithm sentence until it carries a verb, an input, and an output. That is the work that turns a 3 into a 4 and a 4 into a 5.