Key takeaways
- AirOps has no official "15 edits" checklist that I could find. This is a practical checklist built from AirOps' refresh and AEO guidance, plus editing checklists from other publishers and Promptwatch Data.
- The fixes that matter most are factual: verified numbers, named sources, complete specs, and answers placed where an AI can lift them.
- Page type affects citation odds. Product pages and landing pages are about 39% of ChatGPT citations (Sep 28 to Oct 4, 2026), so thin editorial filler is a bad bet.
- Freshness is a quality edit. News pages lose half their ChatGPT citations in 36 days, while landing pages and docs last far longer.
- Do the edits in order: facts first, structure second, voice third, technical last.
A note on the source of this checklist
I went looking for an official AirOps document titled "Content Quality Checklist: 15 Edits" and did not find one. What AirOps does publish is a content refresh strategy guide (October 13, 2026), an AEO audit checklist with 48 factors, and a piece on AI workflows for content refreshes. I only saw the table of contents of the refresh guide, so I am not putting words in its mouth.
So treat this as a working checklist in the AirOps spirit: the edits you would run inside an AirOps workflow, or by hand, before an AI-assisted draft goes live. I borrowed ideas from editing checklists by theStacc, The Write Cure and briefer, and checked the page-type advice against first-party data.
Why bland drafts don't get cited
AI drafts fail in a predictable way. They read smoothly, so mistakes and emptiness hide. The Write Cure puts it well: mistakes hide behind smooth sentences, so they slip through when you read for style instead of checking for truth.
There is a second problem. An AI engine answering a question wants something it can quote: a specific number, a plain definition, a step list, a spec. A draft that says "many businesses benefit from a strategic approach" gives it nothing to quote. A draft that says "the Free plan allows up to 3 users, email support weekdays 9 to 6 Eastern" does.
Promptwatch Data backs this up on the page-type side. In its analysis of ChatGPT citation share by content type (Sep 28 to Oct 4, 2026), product pages took 22.4% of classified citations and landing pages 16.8%. Listicles had 13.6%, documentation 7.5%, how-tos 8.7% and comparisons 5.9%. The report's advice is blunt: keep specs, prices and positioning complete and factual, because ChatGPT quotes them directly.
The 15 edits
Run them in this order. Facts first, style last.
Group 1: Facts and claims (edits 1 to 5)
1. Verify every number, date, name and price. Highlight each one and check it against a real source: your product docs, pricing page, release notes. Not your memory, and not the AI that wrote the draft. The Write Cure's example is a good one: "unlimited users and 24/7 support" becomes "up to 3 users, email support weekdays." Both read fine. Only one is true.
2. Replace vague attributions with named sources. "Studies show" and "experts say" get deleted or replaced with a specific source and a link. If you cannot name it, cut the claim.
3. Hunt absolutes. Search for best, only, always, never, guaranteed, completely, 100%. Keep one only if you can prove it. Promised results like "saves 10 hours a week" need evidence or softer wording.
4. Delete anything you can't independently verify. This hurts, because the unverifiable sentence is often the one that sounds best. Cut it anyway.
5. Add one original data point or first-hand observation per major section. This is the edit that most separates a citable page from a rewrite of the top ten results. It can be a screenshot, an internal number, a test result, a customer quote. theStacc's checklist has a whole originality category for this reason.
Group 2: Structure and extractability (edits 6 to 10)
6. Put the direct answer in the first sentence under each heading. Then elaborate. An AI pulling a passage should not need to read three paragraphs of warm-up.
7. Turn question-shaped headings into real questions. "How long does a content refresh take?" matches how people prompt. "Timeline considerations" matches nothing.
8. Convert comparisons and specs into tables. If a section lists options, features or prices, a table is easier to quote accurately than prose. Make cells complete: no "varies" or "contact us" if you can avoid it.
9. Add a step-by-step section for how-to intent. Promptwatch's report recommends step-by-step guides for informational questions, and how-tos were about 10.7% of citations on informational prompts.
10. Merge sections that repeat the same idea. AI drafts pad by restating. If two H2s say the same thing in different words, keep the better one.
Group 3: Voice (edits 11 to 13)
11. Read the first and last paragraphs aloud. Cut stock phrases like "In today's fast-paced world." If it sounds like a template, rewrite it in your own words.
12. Remove hedges and fake transitions. "It's worth noting," "Let's explore," "Furthermore," "Moreover." Replace transitions with an actual logical link, or nothing.
13. Vary sentence openings and length. Check that no three sentences in a row start the same way. Flat rhythm is the most reliable tell of unedited AI text, and readers bounce on it.
Group 4: Technical and freshness (edits 14 and 15)
14. Fix the technical basics. Target keyword in the title, one H2 and the first 100 words. Meta title around 50 to 60 characters, description around 150 to 160. Internal links to related pages. Click every link, including the ones the AI added, since some will be dead or invented. AirOps' AEO audit also covers confirming AI crawler access, which no amount of editing fixes if bots are blocked.
15. Set a refresh date and owner. Quality decays. A page with prices and specs needs a named person and a review date.
Here is why that last one matters. Promptwatch's content half-life data tracked pages ChatGPT cited on July 6, 2026 for 90 days. After 30 days, ChatGPT still cited 78% of landing pages, 76% of docs and 72% of comparisons. News was the outlier: half of news pages were gone after 36 days. Product pages, reviews and how-tos hit half-life at 74 to 77 days. The practical read is that evergreen and reference pages hold, while time-sensitive content needs a faster refresh cycle.
Which edits matter for which page type
Not every page needs all 15 at full intensity. Use the citation mix to decide where to spend effort.
| Page type | Share of ChatGPT citations (Sep 28 to Oct 4, 2026) | Most important edits | Refresh cadence |
|---|---|---|---|
| Product page | 22.4% | 1, 3, 8, 14 | When prices or specs change |
| Landing page | 16.8% | 1, 6, 8, 11 | Quarterly check |
| Listicle | 13.6% | 1, 2, 5, 8 | Before the 60-day mark |
| How-to guide | 8.7% | 6, 9, 13 | Every 2 to 3 months |
| Documentation | 7.5% | 1, 6, 9 | On every release |
| Comparison | 5.9% | 1, 2, 8 | Every 1 to 2 months |
| News | 5.7% | 1, 2, 15 | Treat as short-lived |
| FAQ page | 3.3% | 6, 7 | Quarterly |
On commercial prompts the concentration is stronger. Promptwatch's report shows product pages (26.0%), landing pages (20.6%), listicles (15.0%) and comparisons (7.9%) add up to almost 70% of citations. If you sell something, your product and landing copy are doing most of the work, and a bland landing page is a real loss.
Running the checklist with a tool stack
You can do all of this in a doc and a spreadsheet. Tools just make it faster, and most of them handle different edits.
| Job | Edits covered | Example tools |
|---|---|---|
| Workflow automation for drafts and refreshes | 6 to 10, 15 | AirOps, Slate |
| On-page optimization and scoring | 7, 14 | Clearscope, Surfer SEO, Frase |
| Grammar and tone | 11 to 13 | Grammarly, Writer |
| Tracking whether pages get cited | 15 and the feedback loop | Promptwatch, Profound, Peec AI |
AirOps for the workflow
AirOps is built for the repeatable part: auditing existing pages, pulling signals, and producing structured refresh drafts. Its refresh guide lists the signals worth watching: organic traffic and rankings, AI citation rate, mention rate, content age, and competitor citation gains. A sensible setup is to build the checklist into the workflow so every draft arrives with sources attached and claims flagged for a human to verify.
The weak point of any automated draft is edits 1 to 5. A workflow can flag a number for checking. It cannot know your real pricing. Keep a human on those.
Optimization and editing tools
Content scorers like Clearscope and Surfer SEO help with edit 14 and with coverage of related terms. They do not tell you whether your claims are true.
For voice edits, a grammar and tone checker catches the mechanical stuff.
Slate positions itself as an AI-search-first content operations platform, which puts it in the same neighborhood as AirOps for teams that want drafting and refresh in one place.
Measuring whether the edits worked
Editing without measurement is guessing. After you publish or refresh, you want to know whether the page is being crawled, cited, and driving visits. Tools like Promptwatch can track which of your pages get cited across ChatGPT, AI Overviews and Perplexity, and show the gap between what you published and what AI actually picks up.
If you want to compare more options, the GEO software directory lists dozens of them.
A 10-minute version
Short on time? The Write Cure's quick pass is good, and I would keep it as your minimum:
- Check every number and name against a source (3 minutes).
- Remove or prove every absolute (2 minutes).
- Read the opening and closing aloud (2 minutes).
- Click every link (1 minute).
- Ask whether the reader knows what to do next (2 minutes).
If you have the full hour, add edits 5 to 10, which are the ones that change whether a page is worth citing rather than just whether it is clean.
Common mistakes when applying a checklist
Asking the AI to fact-check itself
It will confirm its own claims with confidence. Check against a source you control.
Polishing voice before fixing facts
A beautifully edited page with a wrong price is worse than a rough page with the right one, because it reads as trustworthy.
Editing once and walking away
The half-life data above is the argument. Pages with prices, specs and availability drift out of date, and Promptwatch's guidance is to refresh listicles and guides before they drop out of citations.
Treating markdown or schema tricks as a shortcut
Promptwatch's markdown in AI search data found markdown files are just 0.05% of AI search citations. Formatting tricks for bots are not a substitute for a page that says something real.
A bigger-picture note on ChatGPT's citation habits
Citation behavior shifts. Promptwatch Data shows ChatGPT's use of Reddit dropped from about 4% to 0.5% of citations on August 14, 2026, and almost half of responses now cite 10 or more unique domains per Promptwatch's unique-domains report. With that many sources per answer, there is room for a well-edited page from a smaller site to get in, but only if it contains something the others do not.
The takeaway
Most bland drafts do not need a rewrite. They need about 15 specific edits: verify the facts, make the answers quotable, cut the filler, and set a date to check again. Start with edits 1 to 5 on your highest-value product and landing pages, since those earn the largest share of citations, then work outward.