A real AI SEO agency can show you its sourcing process, its fact-checking step, and citation data — not just a traffic dashboard. Ask these 12 questions before you sign anything.
- Sourcing and fact-checking — who verifies claims before publish
- Editorial review — a human name attached to every piece
- Measurement — citation and entity data, not vanity traffic
"AI SEO agency" now covers two very different businesses wearing the same label. One builds retrieval-ready content with real sourcing and review. The other renamed its bulk-content mill in 2024 and kept the assembly line running.
Graphite's analysis of Common Crawl data found primarily AI-generated articles made up 49.9% of new web content by Q1 2026, per Graphite's tracking study — unreviewed machine output at scale, exactly the supply a fake agency draws from.
You can't tell the two apart from a sales deck. You can tell them apart by how they answer these questions.
How do you tell a real AI SEO agency from a content mill wearing a rebrand?
A real agency can name its fact-checking step, produce a human editor for any given article, and show citation or entity-coverage data — a content mill can only show you traffic and volume. The gap shows up the moment you ask for specifics.
Content mills sell speed: "50 articles a month." Genuine AI-SEO work sells structure and accuracy, and volume is whatever that process can sustain without breaking review.
Neither approach is inherently fake — the difference is whether a human verifies the model's output before it goes live. Ask which step happens, not whether AI was involved at all.
The question that ends most sales calls fast: "Walk me through what happens between a draft and a published article." A real process has named steps. A content mill has a publish button.
What should you ask about sourcing and fact-checking?
Ask who verifies each factual claim, whether sources are named inline, and what happens when a model invents a statistic. This is where most content mills go quiet.
- "Who checks a stat before it publishes?" — a named role, not "the AI checks itself."
- "What's your process when a model fabricates a source?" — should describe a catch step, not deny it happens.
- "Can I see a piece where you cut a claim you couldn't verify?" — a real editorial process has examples.
Models hallucinate confidently — fabricated "Dr. [Name]" expert quotes are a common failure mode. Ask to see one article's raw draft next to its published version.
What should you ask about editorial review?
Ask who the human editor is, what they check for beyond grammar, and whether they could explain any paragraph if asked cold. A rebranded content mill outsources this step or skips it.
- "Who is the named editor on this account?" — a person, not a department.
- "What's their review checklist?" — banned phrasing, factual spot-checks, voice consistency are good answers; "we proofread" is not.
- "Could that editor defend any sentence in this piece if a client questioned it?" — tests whether review is real or cosmetic.
Ask for one article's full history — draft, edit notes, final. Agencies doing real review have this on file. Agencies that don't will stall or hand you a clean final draft with no trail behind it.
What should you ask about measurement and reporting?
Ask what gets reported beyond traffic — citations, entity coverage, and ranking movement — since traffic alone can't tell you whether content is working. Google's Search Essentials make clear that usefulness and structure, not volume, are what gets rewarded.
- "What do you report monthly besides traffic?" — citation frequency, entity coverage gaps, and query-level movement are the real signals.
- "How do you measure AI Overview or answer-engine citations?" — should describe a tracked process, not "we check occasionally."
- "What's the baseline you set before starting?" — no baseline means no way to prove the work moved anything.
Compare the pitch to what actually gets penalized. Google's spam policies target scaled content abuse — output produced primarily to manipulate rankings, not help a reader.
What red flags mean you should walk away?
Guaranteed rankings, no named editor, refusal to show a raw draft, and reporting that's 100% traffic with zero citation data are all reasons to stop the conversation. One alone is a warning; two together is a pattern.
Any agency guaranteeing a specific AI Overview placement or ranking position is overselling. Neither Google nor any answer engine gives an outside party that control, and a confident promise here usually means they're confident you won't check.
- "Can you guarantee we'll appear in AI Overviews?" — honest answer is no, with a plan instead.
- "What happens if an error is caught after publish?" — should have a correction process, not silence.
- "Can we see three client examples with before/after data?" — or the offer collapses to a pitch deck.
| Signal | Genuine AI-SEO workflow | Content mill in disguise |
|---|---|---|
| Sourcing | Named sources, fact-check step | Unverified or invented stats |
| Editor | A person you can name and question | "AI + light review" |
| Reporting | Citations, entities, query movement | Traffic and word count only |
| Guarantees | None on placement; tracked progress instead | Promised rankings or citations |
| Draft trail | Available on request | Final copy only, no history |
What should a legitimate deliverable actually look like?
A real AI SEO deliverable is a structured, sourced article with a named editor, plus a report tying it to citation or ranking movement — not a folder of unreviewed drafts. Ask to see the format before you sign, not after.
Structure matters as much as the writing: direct answers under question-form headings, named entities instead of vague categories, and claims attributed to checkable sources. Our breakdown of what AI SEO services actually include covers this in practice, and AI SEO vs traditional SEO covers why the fundamentals underneath it haven't changed.
Pitched on heavy scale — dozens of articles a month? Check whether that pace allows real review; our note on SEO automation covers where automation helps versus replaces the human. Also weighing Korea entry? See our playbook on selling in Korea.
Frequently Asked Questions
Is it a bad sign if an agency uses AI to write content?
No — almost every agency does now, and Graphite's tracking puts primarily AI-generated content at roughly half the web as of Q1 2026. The question isn't whether AI was involved. It's whether a named human fact-checked and edited the output before it published, and whether they can show you that step.
How many client examples should an AI SEO agency show before I sign?
At least three, with before-and-after data on citations, rankings, or entity coverage — not just traffic screenshots. If an agency can only describe results verbally without documentation, treat that as a gap, not a formality you can skip.
Should I ask to see a raw AI draft before editing?
Yes. Comparing a raw draft to the published version shows you exactly what the editorial process actually catches — fabricated stats, invented quotes, generic claims. An agency confident in its review process will show you this without hesitation.
What's a reasonable content volume for genuine AI-assisted SEO work?
It depends on team size, but volume claims should scale with a visible review process, not replace it. If an agency promises 50+ articles a month with a two-person editorial team and no named fact-checking step, the math doesn't support real review at that pace.
Can an AI SEO agency guarantee I'll appear in AI Overviews or ChatGPT answers?
No legitimate agency can, because none of them control the selection logic those systems use. What a real agency can offer is a tracked baseline, structural changes tied to citation-friendly formatting, and monthly reporting on whether citations are actually increasing.
Vetting an agency shouldn't take a law degree — it should take these 12 questions and five honest minutes on a call. If you want a second opinion on a proposal already on your desk, get a free audit and we'll tell you what it's missing.
Last updated: September 2026
Stay Ahead of the Curve
Subscribe to get the latest insights on AI SEO, automation, and high-performance web systems.