Adaptive Learning Paths

“Adaptive” gets used to describe two very different things: software that quietly reroutes a student without telling them, and software that notices a struggle and suggests a next step a student or instructor can accept or ignore. This lesson builds the second kind. It tracks real LearnDash progress signals and turns them into a plain recommendation, never a forced redirect or an automatically relocked lesson.

Step 1: Capture progress signals from real hooks

learndash_quiz_completed fires an action with $quizdata (array) and $user (WP_User) after a quiz is marked complete. learndash_lesson_completed fires with $lesson_data (array) after a lesson is marked complete. Both are real, documented action hooks.

add_action( 'learndash_quiz_completed', function ( $quizdata, $user ) {
	error_log( print_r( $quizdata, true ) );
}, 10, 2 );

Once confirmed, record the signal:

// File: lms-tutor/adaptive-paths.php
add_action( 'learndash_quiz_completed', function ( $quizdata, $user ) {
	$course_id = isset( $quizdata['course'] ) ? (int) $quizdata['course'] : 0;

	if ( ! $course_id ) {
		return;
	}

	update_user_meta( $user->ID, "_lms_tutor_last_quiz_{$course_id}", array(
		'percentage' => $quizdata['percentage'] ?? null,
		'pass'       => $quizdata['pass'] ?? null,
		'time'       => time(),
	) );
}, 10, 2 );

Step 2: Gather progress and remaining steps for a recommendation

// File: lms-tutor/adaptive-paths.php
function lms_tutor_gather_progress( int $user_id, int $course_id ): array {
	$last_quiz = get_user_meta( $user_id, "_lms_tutor_last_quiz_{$course_id}", true );
	$steps     = learndash_get_course_steps( $course_id );

	$completed = array();
	foreach ( $steps as $step_id ) {
		if ( learndash_is_lesson_complete( $user_id, $step_id, $course_id ) ) {
			$completed[] = $step_id;
		}
	}

	return array(
		'last_quiz'  => $last_quiz ?: null,
		'remaining'  => array_values( array_diff( $steps, $completed ) ),
	);
}

learndash_get_course_steps( $course_id ) returns every lesson and topic in the course, which combined with a completion check gives a real, current picture of what’s left, not a guess based on enrollment date alone.

Step 3: Ask for a recommendation, in a fixed, parseable format

// File: lms-tutor/adaptive-paths.php
function lms_tutor_recommend_next_step( int $user_id, int $course_id ): array {
	$progress = lms_tutor_gather_progress( $user_id, $course_id );

	if ( empty( $progress['remaining'] ) ) {
		return array( 'recommendation' => 'course_complete', 'reason' => 'All steps are complete.' );
	}

	$next_step_id  = $progress['remaining'][0];
	$next_title    = get_the_title( $next_step_id );
	$last_score    = $progress['last_quiz']['percentage'] ?? null;
	$last_pass     = $progress['last_quiz']['pass'] ?? null;

	$prompt = <<<PROMPT
A student's most recent quiz result in this course was: percentage {$last_score}, passed: {$last_pass}.
The next unfinished step in the course is: "{$next_title}".
Recommend exactly one of "continue" or "review", with a one-sentence reason. If the
most recent quiz score suggests the student struggled, recommend reviewing the current
material again before moving on. Respond in exactly this format:

Recommendation: <continue or review>
Reason: <one sentence>
PROMPT;

	$response = wp_ai_client_prompt( $prompt )->generate_text();

	$recommendation = 'continue';
	$reason         = '';

	if ( preg_match( '/Recommendation:\s*(\w+)/i', $response, $m ) ) {
		$recommendation = strtolower( trim( $m[1] ) );
	}
	if ( preg_match( '/Reason:\s*(.+)/i', $response, $m ) ) {
		$reason = trim( $m[1] );
	}

	if ( ! in_array( $recommendation, array( 'continue', 'review' ), true ) ) {
		$recommendation = 'continue'; // Unparseable output defaults to no change in path.
	}

	return array( 'recommendation' => $recommendation, 'next_step_id' => $next_step_id, 'reason' => $reason );
}

An unparseable response defaults to continue, the option that changes nothing about the student’s normal path forward, rather than defaulting to review, which would needlessly insert extra work into a student’s course based on a response the code couldn’t actually confirm.

Step 4: Surface it as a suggestion, never an enforcement

Store the result and show it as a dismissible note in the student’s dashboard, or an instructor’s view of that student’s progress:

// File: lms-tutor/adaptive-paths.php
$rec = lms_tutor_recommend_next_step( get_current_user_id(), $course_id );
update_user_meta( get_current_user_id(), "_lms_tutor_recommendation_{$course_id}", $rec );

Nothing in this lesson calls learndash_process_mark_complete(), changes lesson order, or locks any step. The recommendation is data a student or instructor can act on, or ignore entirely and keep moving through the course exactly as they would have anyway.

Test it: simulate a low quiz score and check the recommendation

wp-env run cli wp eval '
update_user_meta( 5, "_lms_tutor_last_quiz_3", array( "percentage" => 40, "pass" => false, "time" => time() ) );
print_r( lms_tutor_recommend_next_step( 5, 3 ) );
'

Confirm the recommendation leans toward review with a reason mentioning the low score, then repeat with a high, passing score and confirm it leans toward continue.

Recap

Real progress signals, learndash_quiz_completed and learndash_lesson_completed, feed a recommendation function that combines the most recent quiz result with learndash_get_course_steps()’s view of what’s left, and asks generate_text() for exactly one of “continue” or “review” plus a reason. The result is stored as visible, dismissible advice, never a change to what content a student can access or complete. The next lesson builds a similarly human-reviewed layer for grading free-text answers.

Resources & further reading

• learndash_quiz_completed hook reference, developers.learndash.com
• learndash_lesson_completed hook reference, developers.learndash.com
• learndash_get_course_steps() function reference, developers.learndash.com