Environmental stories are some of the most rewarding a documentary page can tell, and some of the easiest to get wrong. The temptation runs in two directions. One is to repeat a company's "sustainable" and "eco-friendly" language as though it were a finding. The other is to reach for doom, with a certainty about the end of everything that the evidence doesn't give. Both are failures of precision, and both are avoidable.
This post is about the first problem in particular: greenwashing, meaning environmental claims that suggest more than they can show. I'll explain what two official sets of guidance say about such claims, why they matter even if you don't sell anything, and how I'd check a script. I'll also share a small linter that catches the wording patterns worth a second look. This post is general information, not legal advice. Advertising and consumer-protection rules differ by country, and I am not a lawyer.
Why a documentary page should care
The rules I'll quote were written for businesses that market products and services. The US Federal Trade Commission's Green Guides, in the section on their scope, say they apply to claims about the environmental attributes of a product, package or service "in connection with the marketing, offering for sale, or sale" of that item. The UK Competition and Markets Authority's Green Claims Code is likewise guidance to businesses about consumer protection law.
So a page that simply tells a story isn't what these documents govern. But they matter to a documentary page for three reasons:
- Sponsored content and affiliate links. If you are paid, or earn a commission, to talk about a product or an organisation's work, your content can become part of its marketing. I'm not offering a legal view of exactly when that happens. If it could apply to you, ask a professional.
- Covering a company's claims. If you repeat a firm's environmental claim as fact, you have made the claim yours in the viewer's mind, whatever the law says.
- They are a good checklist. They were built to separate claims that can be backed up from claims that can't, which is the same distinction a good fact-check draws.
What the US Green Guides say
The Green Guides are the FTC's current views on environmental claims. They aren't binding rules, but the Commission can take action under the FTC Act if a claim is inconsistent with them. Four ideas from them are useful for a writer:
- Claims need a reasonable basis. Section 260.2 says that marketers must ensure all reasonable interpretations of their claims are truthful and supported by a reasonable basis, and that in environmental cases this often means competent and reliable scientific evidence.
- Vague claims are a problem. Section 260.4 says unqualified general environmental benefit claims are difficult to interpret, likely convey a wide range of meanings, and may suggest there is no negative environmental impact at all. Because it is highly unlikely a marketer can substantiate all those readings, the Guides say marketers should not make them. Its examples are blunt: a brand name "Eco-friendly" likely conveys far-reaching benefits and is deceptive, whereas "Eco-friendly: made with recycled materials" can be acceptable if the specific statement is clear and prominent and the other conditions are met.
- A true detail can mislead. One example describes a mower advertised as "eco-smart" with "improved fuel efficiency" when the improvement was a tenth of a percent. The claim was technically true and still conveyed a false impression. Another describes a rug labelled "50% more recycled content" because the recycled share rose from 2% to 3%. The guide's principle: don't overstate an attribute, and don't imply a benefit is significant if it is negligible.
- Be specific about what the claim covers. Section 260.3 says a claim should specify whether it refers to the product, the packaging, a service or only a part.
What the UK Code says
The CMA's Green Claims Code sets out six principles. Claims must be truthful and accurate; clear and unambiguous; not omit or hide important relevant information; comparisons must be fair and meaningful; claims must consider the full life cycle of the product or service; and claims must be substantiated. The CMA says the guidance helps businesses understand their obligations under consumer protection law and that it isn't legal advice.
I find the six principles easy to carry over to a script. Each is a question you can ask of a sentence:
- Is it true, and is it exactly true?
- Would two viewers understand it the same way?
- What have I left out that changes the picture?
- Compared with what, and is that a fair comparison?
- Does it consider the whole chain, or only the flattering part?
- Can I point to the evidence?
A claims checklist for a script
This is my own checklist, drawing on the two documents above but not taken from either:
- Attribute claims to their source. "The operator says its harvesting is sustainable" isn't "its harvesting is sustainable." The first is a report of what someone said; the second is a finding you'd need to support yourself.
- Replace general words with specific ones. Say what changed, by how much, over what period, compared with what.
- Give the denominator. "Cut emissions by 30%" means little without the baseline year and the scope.
- Separate claims from measurements. A pledge for 2035 isn't a result. A company's own report isn't an independent one, and it's fair to say which you're quoting.
- Say what you don't know. "We could not find independent figures" is a legitimate sentence in a documentary.
- Check the seal. The Guides note that a certification or seal that doesn't convey its basis can itself imply a general environmental benefit. If you show a logo, say who awards it and for what.
- Watch your own absolutes and doom. "Every species will vanish" and "no impact" need the same evidence as a company's boast.
A claims linter
The script below splits a script into sentences and flags four things: vague green words ("sustainable", "eco-friendly", "carbon neutral", "natural", "clean" and so on), absolute words ("always", "never", "every", "completely", "100%"), doom words ("irreversible", "collapse", "too late"), and numbers with units that don't carry a [src: ...] tag. It's deliberately simple, and each flag comes with a plain-language question rather than a verdict.
claims_lint.py
"""Flag sentences in a script that need a second look: vague green claims, absolutes, doom language, unsourced numbers.
A number is 'sourced' if its sentence carries a tag like [src: Agency report 2025, table 3].
This tool asks questions; it does not decide what is true.
python claims_lint.py script.txt
"""
import re
import sys
VAGUE = ["eco-friendly", "eco friendly", "eco-smart", "green", "greener", "sustainable", "natural", "planet-friendly",
"carbon neutral", "carbon-neutral", "net zero", "net-zero", "zero impact", "environmentally friendly",
"environmentally-friendly", "clean", "responsible"]
ABSOLUTE = ["always", "never", "every", "all of", "completely", "entirely", "totally", "100%", "no impact", "guaranteed",
"proves", "proven", "the only"]
DOOM = ["doomed", "irreversible", "point of no return", "collapse", "apocalypse", "too late", "catastrophe", "wiped out"]
NUMBER = re.compile(r"\b\d[\d,.]*\s?(%|percent|per cent|tonnes|tons|hectares|acres|km|miles|species|years|degrees)\b|\b\d+(?:,\d{3})+\b", re.I)
SRC = re.compile(r"\[src:[^\]]+\]", re.I)
ADVICE = {
"vague": "vague green claim: say the specific benefit and the evidence (what, how much, compared with what)",
"absolute": "absolute wording: is it literally true in every case? If not, say 'most', 'in this study', or give the range",
"doom": "doom language: state the evidence and the uncertainty, and say what is being done or could be done",
"number": "number without a [src: ...] tag: add the source, the date and what was measured",
}
def sentences(text):
text = re.sub(r"^\s*\[(HOOK|BEAT[^\]]*|CLOSE)\]\s*$", "", text, flags=re.M | re.I)
parts = re.split(r"(?<=[.!?])\s+|\n{2,}", text.strip())
return [p.strip() for p in parts if p.strip()]
def has_term(sentence, terms):
s = sentence.lower()
return [t for t in terms if re.search(r"(?<![\w-])" + re.escape(t) + r"(?![\w-])", s)]
def lint(text):
findings = []
for s in sentences(text):
for kind, terms in (("vague", VAGUE), ("absolute", ABSOLUTE), ("doom", DOOM)):
hits = has_term(s, terms)
if hits:
findings.append({"kind": kind, "terms": hits, "sentence": s})
if NUMBER.search(s) and not SRC.search(s):
findings.append({"kind": "number", "terms": [NUMBER.search(s).group(0)], "sentence": s})
return findings
def report(text):
out = []
for f in lint(text):
out.append(f"[{f['kind']}] {', '.join(f['terms'])}\n {f['sentence']}\n -> {ADVICE[f['kind']]}")
return "\n".join(out) if out else "no flags"
if __name__ == "__main__":
print(report(open(sys.argv[1]).read()))
Here is an invented sample script, written to trip it:
sample_claims.txt
[HOOK]
The wetland is doomed, and every species in it will vanish.
[BEAT: The site]
The reserve covers 4,200 hectares. The operator says its harvesting is sustainable and completely carbon neutral.
[BEAT: The evidence]
Survey counts fell from 310 to 220 breeding pairs between 2015 and 2025 [src: Example Survey Group, annual counts, table 2]. The operator's own report calls the practice eco-friendly.
[CLOSE]
We do not know how much of the decline is local. What would a fair test of the operator's claim look like?
The linter flags six things in it: "doomed" as doom language and "every" as an absolute in the hook; the unsourced "4,200 hectares"; "sustainable" and "carbon neutral" as vague, and "completely" as an absolute, in the claim about the operator; and "eco-friendly" in the operator's report. It leaves the survey sentence alone, because that sentence has a source tag. It also leaves "We do not know how much of the decline is local" alone, which is exactly the sort of sentence I want in a script.
The tests check the sentence splitter, the sample's findings, that a sourced number passes and an unsourced one is flagged, that whole words only are matched (so "greenhouse" and "cleaned" aren't flagged), that a precise sourced sentence passes, and that rewording a claim clears its flag. All 6 pass.
test_claims_lint.py
import unittest
from pathlib import Path
from claims_lint import lint, sentences, report
SAMPLE = (Path(__file__).parent / "sample_claims.txt").read_text()
def kinds(text):
return sorted((f["kind"], tuple(f["terms"])) for f in lint(text))
class Tests(unittest.TestCase):
def test_markers_are_ignored_and_sentences_split(self):
s = sentences(SAMPLE)
self.assertFalse(any(x.startswith("[HOOK]") for x in s))
self.assertEqual(len(s), 7)
def test_sample_findings(self):
found = kinds(SAMPLE)
print(report(SAMPLE))
self.assertIn(("doom", ("doomed",)), found)
self.assertIn(("absolute", ("every",)), found)
self.assertIn(("vague", ("sustainable", "carbon neutral")), found)
self.assertIn(("absolute", ("completely",)), found)
self.assertIn(("number", ("4,200 hectares",)), found)
self.assertIn(("vague", ("eco-friendly",)), found)
def test_sourced_number_is_not_flagged_but_unsourced_one_is(self):
self.assertEqual(lint("Counts fell from 310 to 220 breeding pairs [src: Survey, 2025]."), [])
self.assertEqual([f["kind"] for f in lint("The reserve covers 4,200 hectares.")], ["number"])
def test_word_boundaries(self):
self.assertEqual(lint("The greenhouse was cleaned."), []) # 'green' inside 'greenhouse', 'clean' inside 'cleaned'
self.assertEqual([f["kind"] for f in lint("Green hydrogen is not covered here.")], ["vague"]) # a flag is a prompt, not a verdict
self.assertEqual([f["kind"] for f in lint("It is a green product.")], ["vague"])
def test_plain_specific_sentence_passes(self):
self.assertEqual(lint("The plant cut its water use by a third compared with 2020 [src: Operator report 2025, p. 14]."), [])
def test_wording_can_be_fixed(self):
before = lint("We never harm the river.")
after = lint("In our 2025 monitoring, we found no change in the river's oxygen levels [src: Monitoring report, 2025].")
self.assertEqual(len(before), 1)
self.assertEqual(after, [])
if __name__ == "__main__":
unittest.main()
Limits of the linter
- It over-flags. "Green hydrogen" is a technical term and gets flagged anyway. A flag is a prompt to check, not a verdict.
- It under-flags. A vague claim in words that aren't on the list passes. Numbers without a unit from the list ("310 breeding pairs") aren't caught either.
- A source tag isn't a source. The tool checks that a tag exists, not that it's real or says what you claim. That check is yours, as in the source-checking routine.
- It knows English only, and only the lists I wrote.
The other direction
None of this means retreating into mush. A sentence can be careful and still be strong: "Breeding pairs fell from 310 to 220 over ten years, according to the survey group's own counts" is both precise and arresting. Precision makes a story more powerful, because the viewer who later checks the claim finds it holds. And a page that says what it doesn't know earns more trust than one that is certain about everything.
What I haven't verified
- Whether any rule applies to your page. That depends on your country and on whether your content counts as marketing, which I haven't assessed.
- Laws outside the US and UK, including newer rules on environmental claims in other places.
- How regulators have applied the Green Guides or the Code in specific cases. I read the texts, not the enforcement record.
- The linter's word lists against any real set of environmental claims. They're my own, from the examples in the two documents.
This post is general information, not legal advice. Advertising and consumer-protection rules differ by country, and I am not a lawyer.
What I ran this against
- The claims linter was run with Python 3.13.5 (standard library only) on 17 November 2026. 6 tests pass. The sample script, the survey figures, the reserve and the operator in it are all invented for the example and describe nothing real.
- The linter matches words and patterns. It flags sentences for a human to check; it does not know whether a claim is true, false or misleading. Several of its limits are listed in the post.
- I read the US Green Guides and the CMA's Green Claims Code on 17 November 2026. They are rules and guidance aimed at businesses marketing products and services, and I applied them here by analogy to a page covering environmental topics. I'm not a lawyer, and this is not legal advice.
Sources checked
- US eCFR, Title 16, Part 260, "Guides for the Use of Environmental Marketing Claims" (the FTC's Green Guides): § 260.1 (the guides apply to claims about the environmental attributes of a product, package or service in marketing; they are not binding rules but the FTC can act under the FTC Act if a claim is inconsistent with them), § 260.2 (claims must be supported by a reasonable basis, often competent and reliable scientific evidence), § 260.4 (unqualified general environmental benefit claims are difficult to substantiate; examples on "Eco-friendly", "Greener" and "eco-smart"), § 260.3(c) (overstatement of an environmental attribute): ecfr.gov, read 17 November 2026. The eCFR is not an official legal edition.
- UK Competition and Markets Authority, "Making environmental claims on goods and services" (the Green Claims Code), published 20 September 2021. Its six principles: claims must be truthful and accurate; clear and unambiguous; not omit or hide important relevant information; comparisons must be fair and meaningful; claims must consider the full life cycle of the product or service; and claims must be substantiated. It is guidance for businesses and says it is not legal advice: gov.uk, read 17 November 2026.