The 100-Article ChatGPT Mistake

This is the clearest before-and-after I have on what AI content actually does for a family law firm, so I want to walk it in full: what got built, why it sat dead for a year, and exactly what we did to turn it around.

One family law firm we work with hired a content vendor before they hired us. The vendor did what a lot of vendors now do quietly. It opened ChatGPT and published roughly a hundred articles to the firm’s blog.

The articles were not bad. They were grammatical, on-topic, and completely inoffensive. Read any one of them cold and you would have no complaint. That turned out to be the problem.

For about twelve months, the firm’s search traffic did not move. A hundred new pages, and the line on the chart stayed flat the entire year. This is the whole story of why, and the repair that finally bent the curve.

The setup: a content company with a shortcut

The pitch a content vendor makes to a family law firm sounds like relief, and that is exactly why it works. You are a good attorney who dislikes marketing. We will handle your blog. You will never have to think about it again. Ten posts a month, on schedule, at a price that looks reasonable next to your billing rate. Sign here.

What the firm did not see is what “handle your blog” now means for a lot of vendors. Somebody opens ChatGPT, pastes in a list of family law topics, and drops the output into WordPress. The byline on every one of those hundred posts was the vendor’s own email address, not any attorney at the firm. That detail matters more than it looks, and I will come back to it.

None of this required anyone to lie. The posts were real, they were about family law, and they arrived on the schedule everyone agreed to. The firm was paying for output, output showed up, and everyone assumed output was the same thing as progress. Output and progress are different things, and the chart is where the difference showed up.

A year of flat: what 100 articles bought

Twelve months is a long time to watch nothing happen. The posts indexed. They sat in Google. And the firm’s organic traffic stayed flat across the whole year, indifferent to the pile of pages accumulating beneath it. Nobody found them, because there was no reason for anyone to.

A hundred articles can land with zero impact for a structural reason. When five of those posts each took a swing at the same custody question in slightly different words, they did not add up to five times the authority on that question. They split it. Google saw five thin, overlapping pages competing with each other for the same search, and it had no reason to rank any of them over a competitor’s single page that answered the question completely.

Multiply that across a hundred posts and you get a blog that is busy and invisible at the same time. The firm had paid for a year of content and bought a year of flat. Meanwhile a competitor across town, writing one genuinely useful page a month, was quietly pulling ahead. That competitor’s year is the real cost here, and I will put a number on it later.

The diagnosis: why volume without judgment fails

Google has a name for what the vendor did, and the name tells you exactly where the line sits. Its search spam policies, last updated in May 2026, define scaled content abuse as generating many pages “for the primary purpose of manipulating search rankings and not helping users.” The policy names the exact move: “using generative AI tools … to generate many pages without adding value for users.”

Read that wording slowly, because the nuance is the whole lesson. The policy targets purpose and outcome. Pages built to fill a blog and move rankings. Pages that give a reader nothing they could not get anywhere else. It says nothing about which tool typed the words.

I want to be clear about where the fault lies. It is volume with no judgment behind it: a hundred pages that no human with real family law experience ever shaped, questioned, or localized. ChatGPT is not the villain, and I use it in my own work every week. A machine can produce a grammatical paragraph about child custody in seconds. What it cannot produce is the thing that makes that paragraph worth reading: a real client’s question, the way your county commissioner actually handles a relocation request, the moment a general rule breaks and a practitioner knows why. Strip that out and you are left with a hundred pages of nothing, which is precisely what the policy is written to catch.

The repair: pruning, consolidation, and content worth reading

The fix was unglamorous and came in three moves, none of which involved writing a hundred more posts. We took the blog apart in this order.

First, we pruned. A large share of those posts served no reader and never would, so we deleted them outright. This is the step firms resist, because deleting content you paid for feels like setting money on fire. The money was already spent. Keeping dead pages live does not earn it back; it dilutes the pages that could actually work.

Second, we consolidated. Remember the five thin posts all circling the same custody question. We merged them into a single deep page that answered that question the way a client sitting across the table would want it answered. Where five pages had been splitting the topic’s authority five ways and competing with each other, one thorough page concentrated all of it. Consolidation did more quiet work than any other step.

Third, we rewrote the survivors as content only a practitioner could have produced. That meant specific statutes named by number, the actual local courts the firm’s attorneys appear in every week, and the real questions that come up across the consultation table. Every rebuilt piece carried a named, credentialed attorney’s byline instead of the vendor’s email address. That byline is not a formality; it is the signal to a reader, and to Google, that a real expert stands behind the words. The machine-readable side of authorship, the structured markup that connects a page to a real person, lives in our SEO guide. Here the point is simpler and free. The attorney did the thinking. Put the attorney’s name on it.

If you want the fuller version of what to write once the dead weight is gone, the questions worth answering are laid out in our list of family law marketing ideas. Almost every one of them comes straight out of a consultation.

The recovery curve

Then the line moved. The same property that produced twelve flat months on a hundred AI articles climbed to 5,840 clicks and 436,000 impressions on the way up. Same firm, same practice area, same market. The only variable that changed was that experience and judgment replaced raw output.

I want to be honest about what the repair actually cost, because the honest version is more useful than a tidy invoice. The cleanup itself was not the expensive part. The expensive part was the year. For twelve months this firm’s blog did nothing while a competitor across town spent those same twelve months publishing one real, useful page at a time and building the authority this firm could have been building. Content compounds, which is a fine thing when it works for you and a brutal thing when it works for the other side. In a market where an average divorce case is worth $15,000 to $20,000, a year of ceded ground is not a rounding error. That lost year, not the repair bill, is what the hundred articles truly cost.

The firm was not some unlucky outlier, either. The same pattern turned up over and over in our audit of 351 family law firms: blogs full of generic, unattributed posts doing nothing for the firms that paid for them. On at least eight of those sites, a web agency had put its own name or a web developer’s email on the firm’s blog posts, quietly claiming credit for expertise that belonged to the attorneys. Once you know the shape of the mistake, you start seeing it everywhere.

What AI is actually for in law firm content

Here is the honest answer to the question every firm owner is really asking, which is whether they can just use ChatGPT and skip the rest. AI earns its keep as a drafting tool in experienced hands and fails as a substitute for the experience itself.

Handed to someone with real case experience and the judgment to edit hard, AI speeds up a first draft, untangles a clumsy paragraph, turns a lawyer’s rambling voice memo into a clean outline. That is real value, and I use it that way most weeks. The failure starts the moment you ask it to be the experience instead of assist it. A model has never sat across a table from a terrified parent, never watched a specific judge react to a specific argument, never learned which local rule trips people up. Ask it to replace that and you get a hundred articles and twelve flat months, every time.

For a small firm, the sustainable version of this is not a hundred posts a month. It is one deep, genuinely expert, bylined piece a month, built from a real question a real client asked in a real consultation. One a month, done that way, will outperform a hundred generic posts and keep outperforming them for years, because a good page compounds and a filler page decays. How that content cadence fits alongside paid search in a sane family law marketing budget is its own conversation, and the broader case for owning your content rather than renting attention lives in our content marketing guide. The short version is the one this firm learned the hard way: write less, mean more, and put your name on it.

Common questions about AI content for law firms

Can a family law firm use ChatGPT to write its blog?

Yes, as a drafting assistant in the hands of an attorney who edits it hard, not as a replacement for that attorney. AI is good at speeding up a first draft, tightening clumsy writing, and turning a voice memo into an outline. It is poor at supplying the thing that makes legal content rank and convert: real case experience, local specificity, and the judgment to know when a general rule breaks. Draft with it if you want, but do not publish what it hands you without a practitioner shaping it and a named attorney standing behind it.

Will Google penalize AI-written content?

Google does not penalize content for being written with AI. It penalizes content that exists to manipulate rankings without helping anyone, which its May 2026 spam policy calls scaled content abuse. The line Google draws is about purpose, not tool. A useful page written with AI assistance is fine. A hundred generic pages mass-produced to fill a blog are the exact pattern the policy is built to catch, whatever produced them.

How many blog posts should a small family law firm publish?

Fewer than you have been told, done far better. For most small firms, one deep, genuinely expert, bylined article a month beats ten thin ones, because a thorough page that fully answers a real client question concentrates authority while a pile of overlapping posts splits it. Build each piece from an actual consultation question, and a slow, steady cadence you can sustain will outrun a burst of volume you cannot.

Why did a hundred articles produce no traffic?

Because volume by itself is not a ranking strategy. When many thin posts circle the same question, they compete with each other and split whatever authority the topic could earn, so none of them outranks a competitor’s single thorough page. Stack that across a hundred generic, unattributed posts that no experienced attorney shaped, and you get a blog that is busy and invisible at once. Traffic follows pages worth reading, and a hundred interchangeable ones give a reader no reason to choose them.

What should we do with a blog full of old AI content?

Run the same three steps that turned this firm around. Prune the pages that serve no reader and delete them. Consolidate clusters of thin posts on the same question into one deep page that answers it fully. Rewrite the survivors with real statutes, real local courts, and real client questions, and change the byline from a vendor or web developer to the attorney who actually knows the material. You will end with far fewer pages that each do far more work than the whole pile ever did.