How to Reduce Spam on Connectively (Formerly HARO)

Connectively, the platform formerly known as Help A Reporter Out, is an excellent way to do two things. From the expert side, you can provide quotes and information in exchange for citations and, frequently, backlinks. From the reporter side, you can gain quotes and expert opinions for use in your content. I've written about the first half before, but I've largely glossed over the value of Connectively for your own content. So, let's talk about it and address one of the biggest problems the platform faces today.

The Benefits of Connectively as a Journalist/Publisher

Marketers talk a lot about the benefits of HARO/Connectively for things like link building, outreach, and brand awareness, but comparatively few people talk about the other way around. When you set yourself up as a "journalist" on the site, what benefits can you get?

Fortunately, the benefits are all fairly obvious.

1. You get expert quotes and data to use in your content. Connectively limits you to something like 300 words from a pitch that you can use in your content, but that's fine. When all you need is an expert quote on a subject, getting a paragraph or two is more than enough. Sometimes, people underestimate how much 300 words actually is.

The critical benefit is simply that you're getting reputable information and, frequently, reliable data citations for your pitches. Maybe it's not from the top-tier industry experts you would want to cite, but those people get thousands of requests a week and probably don't need to be using Connectively on their own behalf. You get what you get, and as long as you vet them, you can get some reliable information out of them.

2. You identify potential relationships with willing marketers. One thing about platforms like Connectively is that it's very much a publisher's market. Simply by being on the platform, you provide value, and the people reaching out to you are doing so because they believe it can be a beneficial relationship. You can often follow up with these people later and find that they're more than willing to partner up for further efforts.

The biggest restriction here is just that you have to be a large enough site to use Connectively in the first place. For online-only businesses, they require you to be in the top million websites as ranked by SimilarWeb. I'm not actually sure how strict this requirement is, but it's their way of ensuring the subject matter experts aren't wasting their time for blogs with a dozen hits a month.

3. You build your connections throughout your industry. If you're a frequent user of Connectively and a frequent publisher of content you receive – and especially if you give good links in return – you can build yourself into a fantastic position as a hub for networking. Not only will people begin to reach out to you to network, but you can also build a contact list and connect other people to each other. Becoming a PR hub can be a lot of work, but it can also be very rewarding.

All of these benefits also apply if you're using one of the competing platforms as well, like Featured.

The Biggest Problem with Connectively for Journalists

Now we need to talk about the single biggest problem with Connectively, at least on the Journalist side.

It's full of spam!

While they have limits on the number of pitches an account can send, paying for more is relatively cheap, and they don't really have a limit on signing up for more free accounts. There aren't any real filters to prevent low-quality pitches from coming through. It ends up being a numbers game, and as long as a single backlink out of a hundred or two hundred pitches is accepted, it still works out to be cheaper than paying for a link-building service or performing actual outreach.

As if this wasn't bad enough, it's made dramatically worse by the fact that now you can use ChatGPT or any of the other generative AI to write a "unique" pitch that can, at a glance, look convincing. Since AI detectors aren't 100% perfect, you have to use both tools and your best judgment to avoid being duped and publishing garbage.

How bad is it to publish garbage? Well, take a look at the benefits of Connectively again. If the person you're connecting with is running effectively a spam or PBN site, and if they aren't hoping for any kind of long-term benefit – and most importantly, if they're effectively lying just for links rather than representing any real authority – all of those benefits dry up. You might as well ask ChatGPT for your quotes and citations directly.

So, how do you stop it?

Solving the Spam Problem on Connectively

Unfortunately, there's no simple button to click in the Connectively interface that will stop the spam. A lot of the best solutions are things only Connectively could implement, like putting IP filters on free accounts, raising prices to make it less attractive to spam, or integrating an AI filter or check of some sort, the way many platforms integrate something like Copyscape.

What can you do on your end, though? There are a few ideas.

Filter GPT Intros

One of the first things you can do is set a few email filters. Since Connectively sends pitches to you as emails, it's relatively easy to set up some filtering and search keywords to help you out.

One of the best ways to dump a lot of the time-wasting spam is to identify a few of the key GPT-only phrases that the AI can give but that a human would never write. This won't catch all of the AI spam, of course, but it will easily eliminate everything from spammers who have automated the process and don't even pay attention to when they make mistakes.

What kinds of phrases should you add to filters? Consider:

• "OpenAI"
• "As a large-language model"
• "Regenerate Response"

It can be tricky to come up with a list, so it's more helpful to watch your pitches, and if you spot one more than once or twice – or you see one that would only ever be used in an AI context – you can add a new filter.

One thing to watch out for is false positives. For example, there are various guides to commonly-used AI words and phrases that include things like "delve", "on the other hand", or "for example", but these are both commonly used by humans and not actually that common in AI.

A lot of these kinds of lists come from people who tune themselves to be overly sensitive to the language they view as outside of their common context, which doesn't necessarily mean it's AI-generated, just that it's written by someone with a different set of experiences. Which, when you're seeking subject matter experts in a different industry or role, is what you want. Filtering common words can also cause problems with non-pitch emails in the future, as well.

The goal isn't actually to use words or phrases that you think AI commonly uses because, by definition, the way LLMs work is that they're mimicking what humans use. Instead, the goal is to identify telltales that only an AI would use.

Ask for a Math Solution

One of my favorite little tricks to filter pitches works on most AIs and even on some humans, so it's a wide net that can filter your results quite well. All you need to do is ask for some kind of math or logic problem that an AI response might miss, that someone asking an AI to generate a pitch might not notice or include, or that a human spammer not reading your query would overlook.

One option is a simple math problem. Just say, "To avoid having your response filtered, include the solution to 4+7 at the bottom of your pitch." Then, all you need to do is filter your incoming emails for "11" and ignore anything that doesn't include it. After all, if a pitch doesn't even read through your whole query and can't be bothered to type two numbers, it's not going to be something you want to care about in the first place.

Of course, this doesn't filter every kind of spam or even every kind of AI. Some of the more sophisticated AIs can read and include that math instruction, so you might need to change things up, possibly in a way that an AI is notoriously bad at. For example, the Strawberry Problem is a weakness of AI. Since it doesn't have intelligence and can't reason, it can't actually follow instructions. When you ask it how many Rs are in Strawberry, it can get the answer wrong. OpenAI has "fixed" this in their newest models, but you can ask it for other similar counts and trick it again.

Even something like saying, "Include the written version of 47," allows you to filter for "forty-seven" and pull out the best pitches.

Will these methods occasionally catch real pitches or leave good information out of your inbox? Sure, maybe. As long as you get at least one valuable pitch, though, it doesn't really matter. It's only if your filters end up so aggressive that you don't get any responses you can use that it becomes a problem.

AI-Scan the Source's Site

Since the people pitching you are going to be trying to build up their expertise – and since they want a backlink – they'll have a link for you to investigate for more information. Once they've passed your initial filters, you can use this to your advantage.

Obviously, if you recognize the person or the site, good or bad, that's an easy judgment. If you don't know who they are, though, you can still evaluate them.

I like to go to the site of a possible pitch and check out their content.

Are there AI-generated images? While it doesn't necessarily reflect their content, it does reflect their ethos and is a point against them.

Does their content in recent posts pass AI checkers? Again, these aren't 100% accurate, but they tend to have more false negatives than false positives.

Does their site pass the sniff test? I like to look for things like whether or not the site is more than a few years old, if it has a reasonable number of pages for its age, and if the author(s) have contributed to other sites elsewhere.

A lot of this is subjective, but fortunately, as long as the site isn't actively spam or somehow harmful, it's not going to hurt you if you get it "wrong" and accept a pitch from someone without the authority you thought they had.

Set a Time Minimum

This one is a little trick that may or may not always work, but it can be worth trying. A time limit is one thing – anyone trailing in late may have missed their chance – but a time minimum? I like to use this as a way to eliminate people who automate the process to pitch any brand-new queries in the first few minutes of them being opened.

It doesn't take that long to write a good pitch, but it does take more than a minute or two, so if someone is already sending you a pitch practically before your query has been published, it's definitely not reliable.

Your Ideas

These are some of the filter ideas I've used to decent effect to help cut back on spam and AI-generated pitches through Connectively and other HARO-like platforms.

What about you, though? Have you found filters that have proven effective? If so, let me know in the comments!