Table not detected, or detected wrong

import pdfplumber

def why_no_table(page, settings=None):
    """Explain what the table finder saw, step by step."""
    if not page.chars:
        return "no text on the page: scanned image? run OCR first"
    finder = page.debug_tablefinder(settings or {})
    v = [e for e in finder.edges if e["orientation"] == "v"]
    h = [e for e in finder.edges if e["orientation"] == "h"]
    counts = f"{len(v)} vertical / {len(h)} horizontal edges, {len(finder.intersections)} intersections, {len(finder.tables)} table(s)"
    if finder.tables:
        return counts + " -> table found"
    if not v and not h:
        return counts + " -> nothing ruled: use the text or explicit strategy"
    if not v:
        return counts + " -> columns are not ruled: vertical_strategy='text' or explicit lines"
    if not h:
        return counts + " -> rows are not ruled: horizontal_strategy='text' or explicit lines"
    return counts + " -> lines do not meet: raise intersection/join/snap tolerance"

for name in ["invoice", "bank-statement", "borderless-table", "scanned-invoice"]:
    with pdfplumber.open(f"samples/{name}.pdf") as pdf:
        print(f"{name:17} {why_no_table(pdf.pages[0])}")

invoice           6 vertical / 9 horizontal edges, 54 intersections, 1 table(s) -> table found
bank-statement    0 vertical / 24 horizontal edges, 0 intersections, 0 table(s) -> columns are not ruled: vertical_strategy='text' or explicit lines
borderless-table  0 vertical / 0 horizontal edges, 0 intersections, 0 table(s) -> nothing ruled: use the text or explicit strategy
scanned-invoice   no text on the page: scanned image? run OCR first

Symptom → cause → fix

Every row below was reproduced on one of the site’s sample PDFs or on the FBI table from pdfplumber’s test suite. Follow the link for the figure and code.

Only horizontal rules

Bank statements, invoices and ledgers often rule the rows but not the columns. The finder gets plenty of horizontal edges and no vertical ones, so there is nowhere for them to cross:

table edges15words115x — · top —

Give the columns another source of edges: "vertical_strategy": "text", or explicit x positions taken from the header row. The tutorial walks through both and shows why explicit columns are more reliable.

A word split across columns

The text strategy turns any x position where at least three words line up (min_words_vertical, default 3) into a column boundary. In a long list, three descriptions that happen to have a word starting at the same x are enough:

import pdfplumber

with pdfplumber.open("samples/bank-statement.pdf") as pdf:
    page = pdf.pages[0]
    settings = {"vertical_strategy": "text", "horizontal_strategy": "lines"}
    print(page.extract_table(settings)[2])           # 'Lark' | 'spur Robotics'
    settings["min_words_vertical"] = 5               # demand stronger alignment
    print(page.extract_table(settings)[1])

['2026-10-02', 'Payroll — Lark', 'spur Robotics', '', '3,820.00', '6,173.73']
['2026-10-02', 'Payroll — Larkspur Robotics', '', '3,820.00', '6,173.73']

Requiring five aligned words removed the phantom boundary here, but the header row dropped out of the table at the same time, and from 7 a real column disappears. Check the header row after every change, or place the columns explicitly instead.

Doubled letters in cells

When a header is “bolded” by drawing it twice, slightly offset, both copies land in the cell:

import pdfplumber

with pdfplumber.open("samples/framed-table.pdf") as pdf:
    page = pdf.pages[0]
    print(page.extract_table()[0])                  # header drawn twice (fake bold)
    print(page.dedupe_chars().extract_table()[0])   # remove the overprinted copies first

['AAsssseett', 'BBoorrrroowweerr', 'DDuuee bbaacckk']
['Asset', 'Borrower', 'Due back']

dedupe_chars() returns a page without the overprinted copies, and every method works on that page as usual. See missing spaces and doubled letters for the tolerance options.

Try it on your table

Pick a sample with a known problem, or open your own PDF. The Lab shows the vertical and horizontal edge counts and suggests which setting to change:

Run pdfplumber to see extract_text() for this page.

table_settings =   # defaults